One article to understand: the past, present and future of swarm robots

One article to understand: the past, present and future of swarm robots

Cluster robot (swarm robotics) involving large cluster robot design, construction and deployment, they are able to coordinate with each other and collaborate to solve the problem or perform a task. Swarm robots are inspired by natural self-organizing systems, such as social insects, fish or bird flocks, which are all emergent collective behavior based on simple local interaction rules [1][2] . Generally speaking, cluster robots extract engineering principles from the research of these natural systems to construct multi-robot systems with comparable capabilities . In this way, the cluster robot aims to build a system that is more robust, fault-tolerant, and more flexible than a single robot, and can better adjust its behavior to adapt to environmental changes.

Swarm robots, as an application of swarm intelligence [3][4] (that is, the calculation model of collective and self-organizing behavior [5][6]) , have incubated several successful optimization algorithms, which are widely used in the telecommunications industry [7] and simulation and prediction of crowd behavior [8] . However, people quickly discovered that to achieve group behavior in robots, it is more than just applying swarm intelligence algorithms to existing robot platforms.

In fact, researchers need to completely rethink traditional robot functions, such as perception, control, positioning, and the design of the robot platform itself. In the past two decades, researchers in swarm robots have made significant progress. They have provided proof of concept for the potential of swarm robots, and they have also enabled researchers to better understand how complex behaviors occur in nature. Nevertheless, it is still full of challenges to translate this research into practice, which requires researchers to properly solve it.

In fact, so far, only a few experiments have successfully demonstrated a large number of autonomous self-organizing robots, and the practical application of cluster robots is still blank . Researchers need more research to establish relevant theories and practice, so as to bring swarm robots from the laboratory into the real world.

The follow-up content of this article is organized as follows. After a brief introduction to the history of the swarm robotics field, the researchers summarized the main lessons learned during its pioneering phase, analyzed the main open challenges, and gave several innovative and promising Examples of research directions. Finally, by showing the application scenarios that cannot be solved by a single robot, or the application scenarios that cannot be solved by multiple robots controlled by traditional centralized methods, the researchers put forward the most likely application fields of cluster robots and evaluated their impact on selected industries. Potential impact.

1. An overview of the history of swarm robots

In the past 20 years, swarm robots have evolved from a small field initiated by several studies with clear biological inspiration [9-12] to a mature field involving many laboratories and researchers around the world. By using Google Scholar to search, the term “cluster robot” first appeared in 1991, and its use is very limited. This situation lasted until 2003, after which the use of the word began to grow substantially. Similarly, using SCOPUS to search will also get a similar growth trend (as shown in Figure 1) . These data show that although the foundation of swarm robot research comes from some pioneering works in the 1990s, research in this field did not start to grow significantly until 2000.

One article to understand: the past, present and future of swarm robots

Figure 1. Citation counts for searching for “cluster robots” in Google Scholar and Scopus. The graph shows the trend after 2000.

Initially, research on swarm robots aimed to test the use of stigmergy(The definitions of the concepts used in this article are shown in Table 1)As a means of indirect communication and coordination between robots. Following the pace of early research [9-11] , some researches focusing on object retrieval (foraging [13]; Stick pulling [14]) , clustering [15] and object ranking [16] appeared after 2000 . These studies started by observing the known behaviors of social insects and deploying swarm robots with similar behaviors. In a few cases, swarm robots are used to closely replicate the dynamics observed in biological systems (for example, the gathering of cockroach swarms [17]) , which has created a precedent for a bio-robot hybrid society [18] . In addition, robot clusters are also used as tools to solve biological problems [for example, what is the trail network geometry to find the shortest path between the food source and the nest [19]] .

From 2001 to 2005, the Swarm-bots project funded by the European Commission was the first international project to study the cooperation of swarm robots. In this project, a group of up to 20 robots with self-assembly capabilities (that is, physically connected to each other to form a collaborative structure) is used to study some group behaviors, such as collective transportation, area coverage, and target search [25 , 26] . These robots can play different roles in the cluster over time. This is so far the only example of a self-organizing team composed of robots cooperating to solve complex tasks. The Swarmanoid project (2006-2010) extends the ideas and algorithms in Swarm-bots to a heterogeneous robot cluster composed of three types of robots (flying robots, climbing robots, and ground robots) to perform search and retrieval tasks in collaboration [ 28, 29] .

While successfully demonstrating the swarm robot paradigm, research on hardware miniaturization is expected to deploy hundreds or even thousands of collaborative robots (as shown in Figure 2) . Robots are becoming smaller and simpler, and you can even try to design them on the millimeter scale. However, the miniaturization of hardware and the challenges of integrating enough sensor kits have hindered this development process. Only a few years later, Kilbot [30] , a hardware concept that supports 1,000 robot experiments appeared. The original intention of Kilobots was to provide support for the first demonstration of a large robot cluster designed for structural plasticity [31] (shape formation) . Later, it was used in several successful studies, enabling the cluster robot to have hundreds of robots. Demonstration in the physical environment [32-34] .

One article to understand: the past, present and future of swarm robots

Figure 2. Some robots mainly used for cluster robot research: (a) jasmine [35] (picture from Wikimedia Commons); (b) alice [36] (photo courtesy of Simon Garnier); (c) kilobots [30 ] (Photo courtesy of Massimo Berruti); (d) e-pucks [37]; (e) swarm-bots [26]; (f) swarmoid [29]

Swarm robots are not limited to ground platforms : some recent work considers surface [38] and underwater robots [39] , as well as swarms of drones [40, 41] . Although aquatic technology and underwater technology still need a lot of development efforts to mature, drones have been commercialized and represent a very promising platform in remote sensing applications in different fields. At present, it is only due to the lack of authorized autonomous and collective flight. The legal framework is hindered.

In addition to the hardware platform, how to control the robot cluster has become the main focus of research. So far, many documents have provided reports of different methods, but this is beyond the scope of this article (interested readers can refer to [42-46]) . At present, the main research directions include: developing analysis models of cluster systems to guide the realization of robots [47–49] ; adopting (evolutionary) optimization methods, using minimalist controllers (neural network [50], controllers without calculation [51] , 52], finite state machine [53], or grammar-based controller [54]) to guide the robot; development of design and verification methods [20, 55] . As will be discussed below, the engineering method to define a reliable and effective robot cluster is still at the forefront of current research, and efforts are still needed in this area in the next few years.

One article to understand: the past, present and future of swarm robots

Table 1. Glossary

2. Lessons learned and open issues

Although the ultimate goal of swarm robots is to produce methods and tools that make it possible to deploy robot swarms to solve real-world problems, the current focus is still on the scientific understanding of the mechanisms of these methods and tools. The research of the first two decades has left valuable experience and lessons, and also raised some open problems to be solved.

First of all, the researchers learned that the types of tasks that can be performed by robot clusters are strongly restricted by the limited capabilities of autonomous robots. In order to work in a cluster, each robot must be able to interact and communicate, and be able to recognize peers and their work. This requires customized hardware design and specific sensing, processing and interaction capabilities. The current limitations in robot hardware and control limit the complexity of cluster robot research from two aspects.

On the one hand, specific robots have been developed to solve specific (toy-like) problems (for example, termes [56] and kilobots [30]) . These examples open up new research directions, but reusable components are not always portable to different contexts.

On the other hand, general-purpose robots (alice [19, 57] and e-puck [37]) have been used to generate proofs of concept, which usually handle some similar tasks performed by self-organizing natural systems in the artificial world (for example, foraging for food). [13, 34]) direct conversion tasks.

However, when the hardware is not designed for swarm robots, daily work may become very cumbersome : this is because researchers need to process dozens or possibly hundreds of robots at the same time, which makes simple operations (such as charging or uploading software) change. It’s very lengthy and cumbersome. This often limits the number of robots in the tested cluster and reduces the breadth and importance of the demonstration. Finally, it is worth mentioning that hardware miniaturization will be a key element for experiments using a large number of clusters in the laboratory and many future applications. However, the shrinking of the hardware scale has brought extremely difficult problems, which have not been solved so far [58] .

In order to promote the process of swarm robot research, researchers need to develop tools to make it easier for swarm robot researchers to share results and reproduce experiments. Some general-purpose robotic platforms are very valuable tools. e-puck [37] may be the most used cluster robot platform so far, but when the number of e-puck exceeds 30, research activities will become very complicated and expensive. Kilobot is designed for the research of swarm robots and is widely used, but its capabilities are severely limited, so that a virtualized environment is proposed to increase the possibility of research [59, 60] . Although Crazyflies [61] were not conceived for the research of swarm robots, they are also increasingly used as flying platforms for swarm robot research [41] .

Researchers still need a lot of effort to develop the hardware of cluster robots to meet the needs of the research community. First, researchers must find a good compromise between cost, size, and on-board characteristics to ensure that there are enough sensors and actuators, while maintaining size constraints to allow hundreds of robots in the laboratory experiment of.

Accordingly, a size between kilobot and e-puck (approximately 5 cm in diameter) may be a good compromise. The success of e-puck comes from its modular approach, enabling new sensors, actuators or computing capabilities to be plug-in extensions, but this requires careful design. When processing large amounts of data, especially when mobile robots do not require human intervention (for example, when a wireless charging station is integrated into an experimental environment, or when an electric floor is used for battery-free operation) , it is possible to program and charge multiple robots at the same time Possibility (using kilobots) will greatly simplify experimental activities. A centralized system that can observe the state of robots, move robots, record their data and automate experimental activities will speed up the research process and greatly benefit the world.

Simulation hardware is also a basic field of cluster robot research, but its research problems are similar to those of physical robots. Usually, simulation software is developed from scratch for each new robot cluster demonstration. The development of common simulation tools shared by researchers would be a significant advancement because it can simplify the sharing and comparison of research results. However, in order to design such tools, researchers need to better understand the relationship between simulation and the real world. In robotics, this problem is called the simulation-reality gap [62] , that is, when the robot controller developed by simulation is used in the real world, the difference between the model used in the simulation and the model in the real world will cause performance decline. This problem is particularly important in swarm robots, because many robots must interact with each other [44] , and this problem will be further magnified. Even if these differences cannot be completely eliminated, the ideal robot cluster simulator should ensure that they are at the lowest value.

Among the many available simulation software, ARGoS [63] is outstanding in supporting the research of swarm robots. ARGoS makes it possible for up to 10,000 robots to perform real-time dynamic simulation through clever modular design and the possibility of parallel simulation. In addition, it also includes some of the most commonly used cluster robot models (e-pucks and kilbots) .

Another interesting example is Flightmare [64] . It is a (multi) drone simulator that can render the environment realistically and is very useful for the research of visual navigation and remote sensing. In order to improve experience and develop a tool that can respond to the needs of the swarming robot community (while solving the simulation-realistic gap) , researchers need to solve and improve several areas. For example, researchers will need to find ways to improve the simulation of perception and the simulation of physical (robot-robot and robot-environment) and non-physical (communication) interactions. Simulations should be deployed with different fidelities, allowing users to choose a balance between speed and accuracy.

In most cases, high-fidelity simulations are not mandatory, but their availability will greatly simplify the transition from simulation to reality, supporting extensive testing on real robots. It is also necessary to improve the usability of the simulation, which can be achieved by increasing the speed of the simulation and providing a simpler method of processing and controlling the simulated robot and its deployment environment. The simulation should be highly configurable to respond to the needs of the diverse research community. At the same time, the establishment of a new simulation configuration does not require expert knowledge related to the internal operation of the software. Finally, it is important to integrate the simulation framework with standard robot tools and libraries (for example, ROS) and allow cross-compilation for robot platforms, so that real robots can be used to test the code developed in the simulation without any Change or adjust.

With the right tools, the swarm robot research community needs to provide solutions to design problems. In fact, the second lesson learned is how to solve micro and macro problems. Given that researchers can only directly program the individual robots that make up the cluster (at the micro level) , how to design group behavior (at the macro level) may be the most difficult problem. In order to solve this problem, people have tried many times to propose universal design methods that can be reused in different applications (usually guided by biological inspiration) . They include design patterns [20-21] and automatic design methods [50, 53, 65 ] Etc. (see the glossary in Table 1) .

But at present, these methods are not powerful enough: although they successfully solve relatively simple or limited problems, as the complexity of the problem increases, their limitations quickly appear. A complex task consists of several subtasks, which may require cooperation, and have interdependencies and time constraints [66] . People may try to deploy feasible methods for each subtask to obtain modules that can be built later.

However, this divide-and-conquer method is not sufficient to deploy a usable swarm robot system, because this method ignores many possible interrelationships between tasks, and ignores that these tasks can be further divided and scheduled in some way, resulting in A sub-optimal solution. Researchers need to design methods to solve the complex interrelationships between subtasks through continuous integration and optimization [55] .

In addition, current practice needs to expand the size of the group and seamlessly transition from a small group to a large group. Researchers need to design methods for programming robot clusters without considering the size of the cluster/problem, and these should be determined during configuration.

Finally, performance guarantee is very necessary, but current practice is limited to empirical evaluation of performance statistical indicators, and does not fully solve the problem of performance. On the contrary, researchers need to design a method to provide performance limits, so as to meet the verification and inspection standards, and improve the reliability of robot clusters, especially in applications with hard constraints (such as space applications) .

In order to specifically support the research community, benchmarks are a valuable tool that can measure the progress of research in a quantitative way and can also challenge researchers in increasingly complex tasks (for example, the Robot World Cup [68]) . In order to visually illustrate the type of benchmark required for the progress of swarm robot research, researchers now consider such a resource collection problem (as done in the NASA Swarmathon [69] competition) .

In order to go beyond the current practice, researchers can set the problem to adjust its complexity along several dimensions: adjust the size and topology of the environment to test the ability of the proposed solution to adapt to different problem instances, and expand the performance according to the group size; The number and distribution of projects need to be collected to test the ability to coordinate development resources; and to adjust the type and durability of projects to test the ability to identify and retrieve collaboratively, and to adapt to a dynamic environment.

Information complexity should also be changeable, and this can be achieved through multiple alternatives that allow task execution. This will require the cluster to collect and aggregate information related to the problem and its dynamics, and take collective decisions when needed to optimize task performance. If possible, researchers should use variable constraints to identify multiple interrelated tasks within their time execution (for example, give priority to certain item types to support other types of retrieval) . Researchers must assign clear performance indicators to track progress and support comparisons between different methods. If these benchmarks are proposed and associated with standard tools (including the hardware and simulation discussed above) , an open community will form and thrive, learn from best practices and continuously improve existing results.

The third lesson is to understand some of the attributes (for example, fault tolerance and scalability) that are given to the robot cluster . They are not automatically provided by the cluster, but require careful design. If you want to provide other attributes that the self-organizing robot cluster itself does not confer, such as robustness, flexibility, or adaptability (see Table 1) , then the difficulty is even greater.

People try to design robot clusters with these characteristics through theoretical methods, but they ignore the implementation methods and specific functions of robots in terms of sensors and actuators. Researchers have used mathematical models, abstract particle systems or multi-level systems to perform different behaviors in clusters to prove the above properties (such as aggregation [70], collective movement [71], collective decision-making [20] and pattern formation [72]) .

However, transforming theoretical findings into working robotic systems usually requires thorough reflection, including how to introduce features and constraints that are not considered in the necessary simplified theoretical model, and how to interpret the characteristics of the target application domain. In addition, there are still some key issues that have not received enough attention so far, but they are necessary for the deployment of practical applications. Researchers need security to prevent external attacks, so that the cluster can resist malicious users who try to sneak into and catch the cluster. In order for users to interact with the robot system in a meaningful and easy way, how to command and control the cluster is also extremely important. This also requires a high degree of explanatory power, and it is also necessary to promote the acceptance and trust of users and laypersons in the cluster. Solving these problems will greatly improve the cluster robot technology and accelerate its transition from research to specific applications.

The fourth lesson learned is that researchers must use “bio-inspired tools” very carefully. Getting inspiration from the behavior of social insects or social species is very valuable in many cases, because the properties and behaviors of these natural clusters are the basis of any robot cluster: they are self-organizing and can work universally. The “living evidence” of the facts that they provide feasible solutions to specific problems, such as how a cluster of robots move in a coordinated manner, assign tasks, or make collective decisions. In this regard, researchers need to further promote the contribution of biology to provide new guiding principles, because new insights into swarm intelligence mechanisms will continue to provide information for practitioners of swarm robots. However, researchers need to bear in mind the long-term goal of swarm robotics research, which is to deploy clusters of robots that perform useful tasks in the real world.

Therefore, if you want to associate robot clusters with real-world applications, researchers should design robot clusters in an engineering-oriented approach. Therefore, when the behavior required by the robot cluster is highly related to a specific application, the guidance of only relying on biological inspiration is not obvious. Therefore, researchers should avoid having too much confidence in “bio-inspired tools” and be ready to design special solutions when necessary.

It is also worth noting that although the cooperation between biologists and roboticists is fruitful, this cooperation is often one-way, and robotics gains far more than its contribution to biology. Researchers believe that this situation can be improved. Robot clusters can really help biologists by providing artificial and controllable models to study the effects of embodiment , perception, and action, and provide support for group behavior [19] [73] Necessary individual cognitive requirements. In addition, the possibility of integrating autonomous robots into natural clusters provides new research directions [18][74]-[76] , which are just beginning to be explored.

3. New directions and new problems

In the near future, most research on swarm robots is likely to be devoted to seeking answers to the above open questions. This research is of great significance for the further development of this field and the improvement of technical level. However, there are also some research directions that may bring a greater leap, because they may investigate some completely new methods or fields. Although these areas have been identified as open issues, they have not yet been fully studied.

The researchers first discuss when faced with novel and challenging situations, such as extreme constraints caused by small size and a large number of individuals (III-A) , or heterogeneous swarm robots in hardware/behavior (III-B) or In its own organizational structure (III-C), it provides opportunities for researchers to design and control robot clusters. The researchers next consider new directions for designing robot clusters, either imitating biologically inspired examples of responsiveness and adaptability (III-D) , or using machine learning methods to provide clusters with learning capabilities and improve their performance (III-E) . Finally, the researchers discussed the necessity of further research on robot swarm security (III-F) and human-swarm interaction (III-G) , which are essential for real-world deployment.

A. Miniaturization of hardware

One of the goals of swarm robotics technology is to design and control thousands of simple robots to realize complex behaviors at the cluster level generated by simple individual behaviors and a large number of interactions. One aspect that can maximize the future impact of cluster robots is the development of thousands of micro-robots, the size of which can be reduced to millimeters or even micrometers or nanometers. Such clusters can enter small confined spaces (for example, microfluidic channels and the human body) , manipulate microscopic objects (for example, microplastics or individual cells), and self-organize to provide support for local treatments (for example, targeted drug delivery) .

The research so far has only touched the fur of a field with great potential. However, shrinking the size of robots brings new challenges for cluster robots to provide practical solutions. Micro-robots and nano-robots face different laws of physics from the macro-scale and require new groups of behavior patterns.

The current micro-robots and nano-robots do not use traditional hardware, but are composed of active colloidal particles [77] , software (biological) robots [78] , bacteria-driven nanomachines [79,80] , and even controllable genetic engineering Organism [81] composition. Achieving and controlling group behavior in such systems will require a new paradigm, because the ability to precisely control individual behavior will be forcedly restricted.

In addition, it is extremely challenging to integrate traditional methods of perception and action [82] , and researchers need to rethink the strategies for designing and controlling such clusters. In general, research should focus on control methods that use a few unreliable sensors, little or no computing power, and unreliable behavior [51, 52] . It is also reasonable to design the hardware to exhibit self-organizing properties [83 , 84] , although in this case flexible and adaptive behaviors are more difficult to obtain. In all these cases, guiding self-organization is more valuable than trying to control directly.

B. Heterogeneity

The hypothesis of isomorphism still prevails in the study of swarm robots: all robots are the same and run the same control software, they are all replaceable, and only the individual history of interaction with the (social) environment can lead to a certain specific Expression of behavior. This hypothesis originated from the theoretical model of group behavior, which usually simplified complex phenomena to obtain tractability. In fact, self-organization in homogeneous systems is usually sufficient to explain experimental observations [1] . However, different individuals in a natural cluster may be very different in physiology and behavior, and the characteristics of the individual will affect the response to environmental and social cues [85] . Heterogeneity is considered to be the basis for giving group behavior flexibility, adaptability to new conditions, and resilience to external disturbances. These functions are beneficial to cluster robots, but the heterogeneity is not fully utilized as it should be.

The Swarmanoid project mentioned above proved a possible direction by studying the coordinated group behavior in a physical heterogeneous robot cluster [29] . Other powerful forms of collaboration allow initially homogeneous robots to learn different behaviors, and when this benefits group performance, these forms are linked to specific tasks. However, solving the complexity of self-organizing behavior exhibited by heterogeneous entities is still very challenging, but it is expected to bring tremendous progress to the entire field.

C. Decentralization vs. hierarchical structure

From the beginning, swarm robots adopted a self-organizing paradigm, in which group control is obtained through simple (random) rules, which define the way of interaction between robots and with the environment, without the use of any form of concentration Control or global knowledge. However, in many cases, centralized or hierarchical control can make the design and control of cluster robots easier. In many animal societies, hierarchical systems often coexist with self-organization. This fact may also justify the introduction of some form of hierarchical control [86][87] . Unfortunately, these methods will require the introduction of mechanisms that make the system fragile (single point of failure) and difficult to scale.

The choice of decentralization or hierarchical structure, or how to integrate the two, these issues have not yet been fully studied. In this direction, literature [88] first proposed to create a hybrid system, in which the hierarchical control structure generated by the self-organizing process can dynamically appear in a special way. This is similar to what happens in some clusters, where the process of self-organization leads to the formation of linear hierarchies and the emergence of single-reproductive individuals [87] . Mathew et al. [88] created an infrastructure – middleware (Middleware) , allowing the cluster robot from a purely self-organized control automatically switches to the control level, and then automatically switch back. Although experiments have proved the feasibility of this method [88,89] , how to design the rules that allow the creation of hierarchical control structures as a function of tasks that the swarm robot must perform, and how to achieve the transition from pure self-organizing control to hierarchical control ( And the reverse process) is activated as a function of the task and a function of the environment, and understanding these problems still requires a lot of work.

D. Phase transition and adaptability

In the real-world environment, the main challenge facing swarm robots is to adapt to unexpected events, such as the existence of obstacles and changing atmospheric conditions (such as light, wind, and precipitation) . All these events may prevent the cluster robots from moving forward or performing certain tasks. Under these conditions, the cluster must collectively adjust its behavior and automatically change its strategy. This group ability can be observed in some social animals (groups of chironomid, fish, sheep) . In these species, the interaction between individuals will lead to some collective attributes, which are similar to the attributes of a physical system that is close to a phase transition between two macroscopic states (see Table 1 for terms) , which leads to extreme sensitivity to a few individual behavior changes [90, 91] . Under such conditions, the responses of a few individuals who perceive environmental changes can spread to other members of the cluster, enabling them to effectively deal with predator attacks and other disturbances. This group adaptability not only stems from the special interaction between individuals, but also from the adjustment of the relative intensity of these interactions [92] . The conversion of such features in swarm robots can significantly improve their level of autonomy, which will be a promising research direction.

E. Machine learning of cluster robots

So far, the only prominent application of machine learning in cluster robots is evolutionary algorithms (see the glossary in Table 1) , which are used to develop simple neural controllers to drive the behavior of individual robots in the cluster. However, the latest advances in machine learning, especially the availability of new deep learning technologies, can be used as a means of designing group behaviors, as well as providing additional capabilities for individual robots shared within a group.

So far, these studies have not been recognized by the swarm robot community. As a design method, machine learning has problems of automatic design [44] , additional constraints of online learning through trial and error [93] , and situational rewards and coordination problems. Model-free methods (see Table 1) can be very demanding in terms of computational conditions, although they are very powerful in dealing with complex and unpredictable emergencies that characterize group behavior. Since learning (current) models of group behavior can produce highly-designed individual policies, model-based approaches may be valuable. The combination of the two is the direction that several fields have been exploring at present, and it may also be related to the research of swarm robots.

In addition to designing group behaviors, machine learning and more special deep learning methods can provide individual robots with advanced capabilities to maintain individual and group behaviors. From this perspective, it is very important to find a way to use the information obtained by the cluster to support a more effective interpretation of the world. For example, deep networks represent the most advanced technology for image classification, and image classification is a function required in many applications related to robot clusters. By using multiple robots to observe the existence of the same scene from different angles and at different times, a more accurate and computationally efficient solution can be provided [94, 95] . In order to support this cluster-level operation, researchers also need a lot of work to define the network architecture and learning paradigm.

F. Security

The use of autonomous robots outside the laboratory poses safety issues. Robots need to have security when performing tasks [96] . They should ensure the privacy of the collected data, and they should also be able to resist external attacks initiated by malicious users trying to gain control. In the case of robot clusters, these problems will be more serious [97] . Since there may be hundreds of interacting robots in some scenarios, issues such as entity authentication, data confidentiality, and data integrity will be magnified.

In addition, a small number of malicious robots sneaking into the cluster may cause the work of the cluster to be interrupted [97] . The research on the security of robot clusters is still in its infancy. The initial work is to study how to use traditional (for example, encrypted Merkle tree [98]) and less traditional (blockchain [99]) security methods to increase the security layer or be fully integrated into the control architecture of the robot cluster middle. These initial work can solve problems such as how to keep information privatized in the group [98][100] , how to avoid the interference of malicious robots [101] , and how to combat Sybil attacks [99] . Researchers need to conduct a lot of research to extend these simple, proof-of-concept solutions in order to transplant them into large groups of robots in the real world.

G. Human-cluster interaction

Although the interaction with a single machine/robot has been studied in depth [102] , the interaction with the robot cluster opens up a whole new direction. The main difficulty is that the cluster is self-organizing, so there is no clear entity for humans to establish communication with it. In order to provide the cluster with information about the goal to be achieved or the task to be completed [103] [104] , human-cluster interaction (HSI) is very necessary. Embedding some user-driven robots in the cluster can indirectly control the cluster. Recent studies in several disciplines [92][105][107] show that a few loyalty agents can determine the overall behavior of the cluster. Similar mechanisms represent interesting ways to control the robot cluster, although they may need to introduce the necessary security challenges to avoid a small number of malicious robots from controlling the entire robot cluster. Alternatively, the robot cluster can be directly controlled or manipulated by the user, for example, through gestures [108] [109] or EEG signals [110] .

The user’s direct control of the cluster is complicated because it is very challenging to understand what the cluster is doing. This is due to the large number of interactions within the cluster, which may be difficult for human observers to “read”. Therefore, interpretability is crucial. Possible solutions may be built into the self-organizing mechanism of the cluster so that users can see the current state and goals of the cluster. The interface of group behavior (possibly through augmented reality) can collect and visualize information from the cluster, and the model of group behavior can be integrated to provide predictions that support user actions (for example, by issuing new commands to the cluster) . The design of any HSI solution needs to understand the psychological effects of humans interacting with robot clusters to support interactions that reduce stress [111][112] and increase usability and trust [113] .

4. How to guide research in future applications

The great interest in the research of swarm robots [114-116] so far stems from the expectation that real-life applications based on autonomous robots will be ubiquitous in the near future, and that they will cooperate with each other and with human users to avoid the trap of centralized control. At the same time, considering cooperation scenarios (that is, robots coordinate to complete common tasks) and semi-cooperative scenarios (for example, self-interested robots that benefit from global efficient activity organizations, such as self-driving cars) , the knowledge and practices generated by the research of cluster robots will be the solution to future robots The key to complex coordination issues in applications. Therefore, the researchers firmly believe that advancing the research of swarm robots is not only beneficial to the field itself, but to a large extent, it is beneficial to the fields of robotics, cyber-physical systems, and socio-technical systems.

In this section, the researchers first discuss the general quasi-tests of using robot clusters to solve problems or perform tasks in real-world applications, and then outline the main potential application areas of cluster robots that the researchers believe. This overview is speculative, because real-world applications have not yet appeared. However, by considering different application areas and critically evaluating the specific benefits of the swarm robot approach, researchers will have more choices.

A. General guidelines for swarm robot solutions

In principle, when considering the application of robot clusters to solve real-world problems, the first question is whether robot clusters are indeed the best method. However, this is a very difficult problem, especially considering that swarm robotics is a young subject, and as mentioned above, there are still many unsolved research problems in this field. Therefore, current practice includes evaluating the applicability of the swarm robot solution based on the expected advantages relative to other solutions, and considering the constraints imposed by the available technology (a notable exception is Kazadi’s work [117, 118], who made it clear The question of whether the robot cluster is the appropriate technology for a given problem is solved; however, his method is still in the proposal stage and has not yet been applied to any real robot cluster practice) . Due to the lack of working methods from problem specification to robot cluster implementation and deployment, the following researchers will discuss some general guidelines that guide the selection of cluster robot solutions when dealing with specific application problems.

The first very general guideline is that only when a single robot solution cannot (effectively) solve the problem, should you consider using a multi-robot system and its extension-robot clusters, because it takes into account the existing technology And application constraints, these systems are either too complex or too demanding. For example, it is not feasible for a single robot to monitor a large area, and the only option may be to use multiple robots at the same time [119] . Another example is the use of drones to explore a large collapsed building in a search and rescue scenario: even in this case, a drone may perform a mission, but due to the limited flight time and the need to fly back to recharge, this may not be enough efficient. In this case, the multi-robot solution can be more efficient through parallel operation [41] .

Once the adaptability of the multi-robot system is established, researchers should consider which control method is most suitable for the issues involved. For example, when coordinating robots in a centralized manner is unrealistic or undesirable [120] , using robot clusters may be the correct approach. In some cases, centralized re-planning can solve the uncertainty of the task and the unpredictability of the environment [121] . However, if there is a strong demand for online identification and adaptability to incidents, it is best to achieve it through a decentralized, self-organizing method. However, even in this case, one should consider if other methods, such as distributed model predictive control [122, 123] , can be used, which may not be the case when it is impossible or difficult to create a simple model to solve the problem and The environment operation of the robot. However, people should consider whether other methods (such as distributed model predictive control [122, 123]) can be used , but when it is impossible or too difficult to create a simple enough model to deal with the problem to be solved and the environment in which the robot will operate, That’s another matter.

Another aspect to consider is whether a given problem can be decomposed into a fixed number of clearly defined tasks that can be completed by a group of robots, and each robot has a specific role, such as an assembly line or a robot soccer [ 68] . If this is not the case, then the swarm robot approach may be feasible. In other words, even if a problem can be better solved with a multi-robot system, this does not necessarily mean that a robot cluster is needed. If the task has no predefined partitions in the subtasks, or the task allows different roles to be assigned to available robots [27 , 29] , the task is more suitable for the latter scheme. Finally, if beneficial cooperation between robots is desired, the swarm robot approach may be the right choice. In fact, the colony robot system can achieve super-linear performance growth through cooperation, which proves that the cost necessary to establish cooperation is reasonable [124] .

B. Applications, needs and future research

Considering these factors, researchers should conduct a rigorous assessment of the potential application areas of swarm robots to determine the specific benefits that the swarm robot method can bring. For example, although service robots are usually not organized in groups, the coordination activities and task assignments performed by each robot are decentralized and self-organized to a certain extent. Nevertheless, the specific task itself may not require coordination or collaboration between robots. Similarly, logistics (such as large warehouses) , self-driving cars, and smart mobility can certainly benefit from the decentralized coordination strategy of cluster robot research.

However, these applications are unlikely to guide future swarm robot research. In contrast, applications such as precision agriculture or infrastructure inspection and maintenance need to deal with unstructured, unpredictable environments (often covering a wide range) , and they can benefit from the parallelization and collaboration of robot clusters. For example, early identification of disease outbreaks in farmland requires information sharing between robots in order to form a global model from a coupled local perspective, supporting appropriate responses and better strategic planning [95, 125] . Similarly, reliable identification of defects in large-scale infrastructure requires efficient search capabilities, and this capability can be best achieved by clustering [126] . Both precision agriculture and infrastructure inspections occur in a static environment (farmland or infrastructure to be inspected) .

Nonetheless, decentralization and self-organization can improve efficiency (thanks to parallel and coordinated operations) and accuracy (thanks to group-aware adaptive strategies, which allow to react to perceived incidents and determine optimal tasks Plan to maximize the likelihood of all relevant features being carefully observed) . In this regard, future research should focus on the strategy of understanding complex features through information fusion between a variety of potentially heterogeneous robots. In addition, researchers need to design targeted interventions and operational capabilities (such as fruit harvesting or maintenance) to provide new opportunities for decentralized cooperative activities.

Defense agencies all over the world are looking for the application of robot clusters, and found that systems that cannot be easily shut down are very attractive [127] . Systems that are fault-tolerant to external attacks can support operations in adversarial settings, especially when the robot is replaceable and, to a certain extent, discardable. However, in this regard, the human factor is still inevitably at the center. Therefore, national defense applications need to consider human beings in the environment, and advanced HSI strategies will be the key to effective deployment [113] .

In addition, the highest level of safety and protection needs to be achieved to ensure that the robot cluster will not lose control or be maliciously captured [96] . Similar aspects are also important in other application fields such as civil defense. These fields need to face natural or man-made disasters and require agile robots that can handle emergencies without relying on external infrastructure or reliable maps. The threshold for this type of application is very high, because the robot cluster needs to ensure the highest performance and reliability to ensure that all victims are rescued.

Space tasks introduce other limitations of robot applications, which may be successfully resolved by swarm robots. In space, due to the influence of cosmic radiation on modern CPUs, the computing power of computers is still limited [128] . Therefore, a cluster of robots with limited computing power may be a better design choice than a single, more powerful robot [69, 129, 130] . Robots launched into space are not easy to repair or replace. A cluster robot dedicated to a redundant system would be a good solution. The failure of a single robot in the cluster will only cause a slight decrease in the performance of the cluster.

Finally, in space, it may be extremely expensive or even impossible to establish external infrastructure to support the coordination of robots. This is also a typical situation that a robot cluster can effectively handle. Therefore, space agencies such as NASA and ESA have begun to be interested in cluster technology, such as the aforementioned activities such as the Swarmathon competition [69] and research on the control of microsatellite swarms [130] . The necessary autonomy of group systems is a huge challenge brought by space applications, and it requires that reliable and continuous manual intervention cannot be relied on.

Robot clusters also have good development potential in the entertainment industry. There are already some examples of drones performing light shows outdoors and indoors [131] , however, this is usually based on centralized pre-arranged flight paths. Similarly, other attempts to develop multi-robot entertainment systems also rely on some centralized control solutions to finely control the system [132, 133] . If you consider a decentralized approach, especially if users can actively participate in entertainment activities by participating in robot clusters, and change their dynamics according to location, movement and even emotions, there may be new opportunities [134] . In this case, future research can test the new model of HSI, which can then be used for reference by other application fields. For example, researchers can imagine various HSI interfaces, from wearable devices [135] , augmented and virtual reality [136] to brain-computer interaction modes [110] .

Finally, swarms of nanorobots may become high-tech tools for precision medicine in the future, enabling targeted interventions in the human body, such as minimally invasive surgery or multiple therapies delivered directly to cancer cells [137, 138] . However, coordinating a large number of robots with extremely limited computing power and communication capabilities will bring the swarm robot approach to its limit and require the development of new conceptual tools, as well as microscopic hardware or biological robotic equipment [58] .

In general, there is no doubt about the relationship between the needs of the potential application areas of swarm robots and future research challenges. Therefore, the researchers envisioned close cooperation between researchers and stakeholders from different application fields, who can provide examples to promote new developments and contribute to setting the research agenda for cluster robots in the coming years.

Five, summary

The design and implementation of effective robot clusters is one of the biggest challenges facing robotics and one of the most promising research directions. This statement has been confirmed [116] . In this article, the researchers briefly summarized the status quo of swarm robotics technology, and identified the most promising research directions and major open issues that the researchers believe. However, researchers should believe that major advances in swarm robotics technology will inevitably make progress outside the field. For example, new materials, bio-hybrid solutions, and new methods of storing and transmitting energy will help solve some of the current problems related to robotic cluster hardware.

The development of artificial intelligence technology, especially the development of distributed learning algorithms that only require limited computing resources and can work with the CPU of small and cheap robots, will gradually increase the autonomy of the robot community. The cluster must ensure interpretability, which is now a major problem in the entire field of robotics and artificial intelligence. In other words, users will need to be able to understand the decision-making process without a detailed understanding of the underlying mechanisms-this is an important requirement for ensuring the acceptability of new smart technologies and fostering trust in them, thus providing large-scale real-world applications The deployment creates conditions. Although these problems have been more widely solved in the field of artificial intelligence, their complexity may be increased by a large number of autonomous entities and a large number of interactions between them, which is precisely the typical feature of swarm robot systems.

If researchers can overcome these challenges, swarm robots are expected to successfully enter the real world from the laboratory within ten years. This transformation will not happen immediately, but will gradually involve more and more application areas, thereby identifying new challenges and creating demand for emerging technology solutions, thereby driving research and innovation in the coming years.


[1] S. Camazine, J.-L. Deneubourg, N. R. Franks,J. Sneyd, G. Theraulaz, and E. Bonabeau, Self-Organization in Biological Systems. Princeton,NJ, USA: Princeton Univ. Press, 2001.

[2] D. J. T. Sumpter, Collective Animal Behavior.Princeton, NJ, USA: Princeton Univ. Press, 2010.

[3] E. Bonabeau, M. Dorigo, and G. Theraulaz, Swarm Intelligence: From Natural to Artificial Systems. New York, NY, USA: Oxford Univ. Press, 1999.

[4] M. Dorigo and M. Birattari, “Swarm intelligence,” Scholarpedia, vol. 2, no. 9, p. 1462, 2007.

[5] M. Dorigo, “Ant colony optimization,” Scholarpedia, vol. 2, no. 3, p. 1461, 2007.

[6] M. Dorigo,M. A. M. de Oca, and A. P. Engelbrecht, “Particle swarm optimization,” Scholarpedia, vol. 3, no. 11, p. 1486, 2008.

[7] G. Di Caro and M. Dorigo, “AntNet: Distributed stigmergetic control for communications networks,” J. Artif. Intell. Res., vol. 9, pp. 317–365, Dec. 1998.

[8] M. Moussaïd, D. Helbing, and G. Theraulaz, “How simple rules determine pedestrian behavior and crowd disasters,” Proc. Nat. Acad. Sci. USA, vol. 108, no. 17, pp. 6884–6888, 2011. [Online]. Available:

[9] R. Beckers, O. E. Holland, and J. L. Deneubourg,“From local actions to global tasks: Stigmergy and collective robotics,” in Proc. 4th Int. Workshop Synthesis Simulation Living Syst. Artif. Life IV, R. A. Brooks and P. Maes, Eds. Cambridge, MA, USA:MIT Press, 1994, pp. 181–189.

[10] O. Holland and C. Melhuish, “Stigmergy, self-organization, and sorting in collective robotics,” Artif. Life, vol. 5, no. 2, pp. 173–202,Apr. 1999.

[11] C. R. Kube and H. Zhang, “Collective robotics:From social insects to robots,” Adapt. Behav.,vol. 2, no. 2, pp. 189–218, Sep. 1993.

[12] C. Kube and E. Bonabeau, “Cooperative transport by ants and robots,” Robot. Auton. Syst., vol. 30,no. 1, pp. 85–101, 2000. [Online]. Available:

[13] M. J. B. Krieger, J.-B. Billeter, and L. Keller,“Ant-like task allocation and recruitment in cooperative robots,” Nature, vol. 406, no. 6799,pp. 992–995, Aug. 2000.

[14] A. J. Ijspeert, A. Martinoli, A. Billard, and L. M. Gambardella, “Collaboration through the exploitation of local interactions in autonomous collective robotics: The stick pulling experiment,” Autonom. Robots, vol. 11, no. 2, pp. 149–171,2001.

[15] W. Agassounon, A. Martinoli, and K. Easton,“Macroscopic modeling of aggregation experiments using embodied agents in teams of constant and time-varying sizes,” Auton. Robots,vol. 17, nos. 2–3, pp. 163–192, Sep. 2004.

[16] M.Wilson, C. Melhuish, A. B. Sendova-Franks, and S. Scholes, “Algorithms for building annular structures with minimalist robots inspired by brood sorting in ant colonies,” Auton. Robots, vol. 17, nos. 2–3, pp. 115–136, Sep. 2004.

[17] S. Garnier et al., “The embodiment of cockroach aggregation behavior in a group of micro-robots,”Artif. Life, vol. 14, no. 4, pp. 387–408, Oct. 2008.

[18] J. Halloy et al., “Social integration of robots into groups of cockroaches to control self-organized choices,” Science, vol. 318, no. 5853, pp. 1155–1158, 2007. [Online]. Available:

[19] S. Garnier, M. Combe, C. Jost, and G. Theraulaz,“Do ants need to estimate the geometrical properties of trail bifurcations to find an efficient route? A swarm robotics test bed,” PLoS Comput. Biol., vol. 9, no. 3, Mar. 2013, Art. no. e1002903.

[20] A. Reina, G. Valentini, C. Fernández-Oto,M. Dorigo, and V. Trianni, “A design pattern for decentralised decision making,” PLoS ONE, vol. 10, no. 10, Oct. 2015, Art. no. e0140950.

[21] G. Valentini, Achieving Consensus in Robot Swarms: Design and Analysis of Strategies for the Best-of-n Problem (Studies in Computational Intelligence). Cham, Switzerland: Springer, 2017.

[22] P.-P. Grassé, “Reconstruction of the nest and interindividual coordination in Bellicositermes natalensis and Cubitermes sp. The Stigmergy Theory: An attempt to interpret the behavior of constructor termites, ”Social Insects, vol. 6, no. 1, pp. 41–80, Mar. 1959.

[23] G. Theraulaz and E. Bonabeau, “A brief history of stigmergy,” Artif. Life, vol. 5, no. 2, pp. 97–116, 1999, doi: 10.1162/106454699568700.

[24] S. Garnier, J. Gautrais, and G. Theraulaz,“The biological principles of swarm intelligence,”Swarm Intell., vol. 1, no. 1, pp. 3–31, Oct. 2007.

[25] M. Dorigo et al., “Evolving self-organizing behaviors for a swarm-bot,” Auton. Robots,vol. 17, nos. 2–3, pp. 223–245, Sep. 2004.

[26] F. Mondada, L. M. Gambardella, D. FloreanoS. Nolfi, J. Deneubourg, and M. Dorigo, “The cooperation of swarm-bots–physical interactions in collective robotics,” IEEE Robot. Autom. Mag., vol. 12, no. 2, pp. 21–28, Jun. 2005.

[27] S. Nouyan, R. Groß,M. Bonani, F.Mondada, and M. Dorigo, “Teamwork in self-organized robot colonies,” IEEE Trans. Evol. Comput., vol. 13, no. 4, pp. 695–711, Aug. 2009.

[28] M. Dorigo, M. Birattari, and R. O’Grady, “Swarmanoid, the movie,” in Proc. AAAI-11 Video,2011. [Online]. Available:

[29] M. Dorigo et al., “Swarmanoid: A novel concept for the study of heterogeneous robotic swarms,”IEEE Robot. Autom. Mag., vol. 20, no. 4,pp. 60–71, Dec. 2013.

[30] M. Rubenstein, C. Ahler, N. Hoff, A. Cabrera, and R. Nagpal, “Kilobot: A low cost robot with scalable operations designed for collective behaviors,” Robot. Auton. Syst., vol. 62, no. 7, pp. 966–975,2014, reconfigurable Modular Robotics. [Online]. Available:

[31] M. Rubenstein, A. Cornejo and R. Nagpal, “Programmable self-assembly in a thousand-robot swarm”, Science, vol. 345, no. 6198, pp. 795-799, 2014, [online] Available:

[32] G. Valentini, E. Ferrante, H. Hamann and M. Dorigo, “Collective decision with 100 kilobots: Speed versus accuracy in binary discrimination problems”, Auton. Agents Multi-Agent Syst., vol. 30, no. 3, pp. 553-580, May 2016.

[33] I. Slavkov et al., “Morphogenesis in robot swarms,” Sci. Robot., vol. 3, no. 25, 2018, Art. no. eaau9178. [Online]. Available: https://

[34] M. S. Talamali, T. Bose, M. Haire, X. Xu, J. A. R. Marshall, and A. Reina, “Sophisticated collective foraging with minimalist agents: A swarm robotics test,” Swarm Intell., vol. 14, no. 1, pp. 25–56, Mar. 2020.

[35] S. Kornienko, O. Kornienko, and P. Levi, “Minimalistic approach towards communication and perception in microrobotic swarms,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst., Aug. 2005, pp. 2228–2234.

[36] G. Caprari and R. Siegwart, “Mobile micro-robots ready to use: Alice,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst., Aug.2005, pp. 3295–3300.

[37] F. Mondada et al., “The e-puck, a robot designed for education in engineering,” in Proc. 9th Conf. Auton. Robot Syst. Competitions, 2009, pp. 59–65.

[38] M. Duarte et al., “Evolution of collective behaviors for a real swarm of aquatic surface robots,” PLoS ONE, vol. 11, no. 3, Mar. 2016, Art. no. e0151834.

[39] P. Zahadat and T. Schmickl, “Division of labor in a swarm of autonomous underwater robots by improved partitioning social inhibition,” Adapt. Behav., vol. 24, no. 2, pp. 87–101, Apr. 2016.

[40] G. Vásárhelyi, C. Virágh, G. Somorjai, T. Nepusz, A. E. Eiben, and T. Vicsek, “Optimized flocking of autonomous drones in confined environments,” Sci. Robot., vol. 3, no. 20, 2018, Art. no. eaat3536. [Online]. Available: https://robotics.sciencemag. org/content/3/20/eaat3536

[41] K. N. McGuire, C. De Wagter, K. Tuyls, H. J. Kappen, and G. C. H. E. de Croon, “Minimal navigation solution for a swarm of tiny flying robots to explore an unknown environment,” Sci. Robot., vol. 4, no. 35, 2019, Art. no. eaaw9710. [Online]. Available: https://robotics.

[42] M. Brambilla, E. Ferrante, M. Birattari, and M. Dorigo, “Swarm robotics: A review from the swarm engineering perspective,” Swarm Intell., vol. 7, no. 1, pp. 1–41, Mar. 2013.

[43] V. Trianni and A. Campo, “Fundamental collective behaviors in swarm robotics,” in Springer Handbook of Computational Intelligence, J. Kacprzyk and W. Pedrycz, Eds. Berlin, Germany: Springer, 2015, pp. 1377–1394, doi: 10.1007/978-3-662-43505-2_71. [44] G. Francesca and M. Birattari, “Automatic design of robot swarms: Achievements and challenges,” Frontiers Robot. AI, vol. 3, p. 29, Mar. 2016. [Online]. Available: https://www.frontiersin. org/article/10.3389/frobt.2016.00029

[45] L. Garattoni and M. Birattari, “Swarm robotics,” in Wiley Encyclopedia of Electrical and Electronics Engineering, J. G. Webster, Ed. 2016, 10.1002/047134608X.W8312.

[46] G. Valentini, E. Ferrante and M. Dorigo, “The best-of-n problem in robot swarms: Formalization state of the art and novel perspectives”, Frontiers Robot. AI, vol. 4, Mar. 2017.

[47] M. Massink, M. Brambilla, D. Latella, M. Dorigo, and M. Birattari, “On the use of bio-PEPA for modelling and analysing collective behaviours in swarm robotics,” Swarm Intell., vol. 7, nos. 2–3, pp. 201–228, Sep. 2013.

[48] A. Prorok, N. Correll, and A. Martinoli, “Multi-level spatial modeling for stochastic distributed robotic systems,” Int. J. Robot. Res., vol. 30, no. 5, pp. 574–589, Apr. 2011.

[49] K. Elamvazhuthi and S. Berman, “Mean-field models in swarm robotics: A survey,” Bioinspiration Biomimetics, vol. 15, no. 1, 2019, Art. no. 015001, doi: 10.1088/1748-3190/ab49a4.

[50] V. Trianni, “Evolutionary swarm robotics,” in Evolving Self-Organising Behaviours in Groups of Autonomous Robots (Studies in Computational Intelligence), vol. 108. Berlin, Germany: Springer-Verlag, 2008.

[51] M. Gauci, J. Chen, W. Li, T. J. Dodd, and R. Groß, “Self-organized aggregation without computation,” Int. J. Robot. Res., vol. 33, no. 8, pp. 1145–1161, 2014.

[52] A. Ozdemir, M. Gauci, S. Bonnet, and R. Groß, “Finding consensus without computation,” IEEE Robot. Autom. Lett., vol. 3, no. 3, pp. 1346–1353, Jul. 2018.

[53] G. Francesca et al., “AutoMoDe-chocolate: Automatic design of control software for robot swarms,” Swarm Intell., vol. 9, nos. 2–3, pp. 125–152, Sep. 2015.

[54] E. Ferrante, A. E. Turgut, E. Duéñez-Guzmán, M. Dorigo, and T. Wenseleers, “Evolution of self-organized task specialization in robot swarms,” PLOS Comput. Biol., vol. 11, no. 8, Aug. 2015, Art. no. e1004273.

[55] M. Brambilla, A. Brutschy, M. Dorigo, and M. Birattari, “Property-driven design for robot swarms: A design method based on prescriptive modeling and model checking,” ACM Trans. Auton. Adapt. Syst., vol. 9, no. 4, pp. 17:1–17:28, 2015.

[56] J. Werfel, K. Petersen, and R. Nagpa, “Designing collective behavior in a termite-inspired robot construction team,” Science, vol. 343, no. 6172, pp. 754–758, 2014. [Online]. Available: https://

[57] S. Garnier, J. Gautrais, M. Asadpour, C. Jost, and G. Theraulaz, “Self-organized aggregation triggers collective decision making in a group of cockroach-like robots,” Adapt. Behav., vol. 17, no. 2, pp. 109–133, Apr. 2009.

[58] M. Sitti, Mobile Microrobotics. Cambridge, MA, USA: MIT Press, 2017.

[59] A. Reina, A. J. Cope, E. Nikolaidis, J. A. R. Marshall, and C. Sabo, “ARK: Augmented reality for kilobots,” IEEE Robot. Autom. Lett., vol. 2, no. 3, pp. 1755–1761, Jul. 2017.

[60] G. Valentini et al., “Kilogrid: A novel experimental environment for the Kilobot robot,” Swarm Intell., vol. 12, no. 3, pp. 245–266, Sep. 2018.

[61] W. Giernacki, M. Skwierczy´nski, W. Witwicki, P. Wro´nski, and P. Kozierski, “Crazyflie 2.0 quadrotor as a platform for research and education in robotics and control engineering,” in Proc. 22nd Int. Conf. Methods Models Autom. Robot. (MMAR), Aug. 2017, pp. 37–42.

[62] N. Jacobi, P. Husbands, and I. Harvey, “Noise and the reality gap: The use of simulation in evolutionary robotics,” in Proc. 3rd Eur. Conf. Adv. Artif. Life. Berlin, Germany: Springer-Verlag, 1995, pp. 704–720.

[63] C. Pinciroli et al., “ARGoS: A modular, parallel, multi-engine simulator for multi-robot systems,” Swarm Intell., vol. 6, no. 4, pp. 271–295, Dec. 2012.

[64] Y. Song, S. Naji, E. Kaufmann, A. Loquercio, and D. Scaramuzza, “Flightmare: A flexible quadrotor simulator,” 2020, arXiv:2009.00563. [Online]. Available:

[65] M. Birattari et al., “Automatic off-line design of robot swarms: A manifesto,” Frontiers Robot. AI, vol. 6, p. 59, Jul. 2019.

[66] B. P. Gerkey and M. J. Matari´c, “A formal analysis and taxonomy of task allocation in multi-robot systems,” Int. J. Robot. Res., vol. 23, no. 9, pp. 939–954, Sep. 2004.

[67] E. Nunes, M. Manner, H. Mitiche, and M. Gini, “A taxonomy for task allocation problems with temporal and ordering constraints,” Robot. Auton. Syst., vol. 90, pp. 55–70, Apr. 2017. [Online]. Available: http://www.sciencedirect. com/science/article/pii/S0921889016306157

[68] S. R. Hedberg, “Robots playing soccer? RoboCup poses a new set of challenges in intelligent distributed computing,” IEEE Concurrency, vol. 5, no. 4, pp. 13–17, Oct. 1997.

[69] L. A. Nguyen, T. L. Harman, and C. Fairchild, “Swarmathon: A swarm robotics experiment for future space exploration,” in Proc. IEEE Int. Symp. Meas. Control Robot. (ISMCR), Sep. 2019, pp. B1-3-1–B1-3-4.

[70] H. Hamann and H. Wörn, “A framework of space–time continuous models for algorithm design in swarm robotics,” Swarm Intell., vol. 2, nos. 2–4, pp. 209–239, Dec. 2008.

[71] T. Vicsek and A. Zafeiris, “Collective motion,” Phys. Rep., vol. 517, nos. 3–4, pp. 71–140, Aug. 2012.

[72] M. Coppola, J. Guo, E. Gill, and G. C. H. E. de Croon, “Provable self-organizing pattern formation by a swarm of robots with limited knowledge,” Swarm Intell., vol. 13, no. 1, pp. 59–94, Mar. 2019.

[73] L. Lei, R. Escobedo, C. Sire, and G. Theraulaz, “Computational and robotic modeling reveal parsimonious combinations of interactions between individuals in schooling fish,” PLOS Comput. Biol., vol. 16, no. 3, Mar. 2020, Art. no. e1007194.

[74] J. Krause, A. F. Winfield, and J.-L. Deneubourg, “Interactive robots in experimental biology,” Trends Ecol. Evol., vol. 26, no. 7, pp. 369–375, 2011. [Online]. Available: http://www. 4711000851

[75] S. Mitri, S. Wischmann, D. Floreano, and L. Keller, “Using robots to understand social behaviour,” Biol. Rev., vol. 88, no. 1, pp. 31–39, 2013, doi: 10.1111/j.1469-185X.2012.00236.x.

[76] F. Bonnet et al., “Robots mediating interactions between animals for interspecies collective behaviors,” Sci. Robot., vol. 4, no. 28, 2019, Art. no. eaau7897. [Online]. Available: https://

[77] H. Xie et al., “Reconfigurable magnetic microrobot swarm: Multimode transformation, locomotion, and manipulation,” Sci. Robot., vol. 4, no. 28, 2019, Art. no. eaav8006. [Online]. Available: eaav8006

[78] S. Kriegman, D. Blackiston, M. Levin, and J. Bongard, “A scalable pipeline for designing reconfigurable organisms,” Proc. Nat. Acad. Sci. USA, vol. 117, no. 4, pp. 1853–1859, 2020. [Online]. Available: https://www.pnas. org/content/117/4/1853

[79] R. D. Leonardo et al., “Bacterial ratchet motors,” Proc. Nat. Acad. Sci. USA, vol. 107, no. 21, pp. 9541–9545, May 2010. [Online]. Available:

[80] K. E. Peyer, L. Zhang, and B. J. Nelson, “Bio-inspired magnetic swimming microrobots for biomedical applications,” Nanoscale, vol. 5, no. 4, pp. 1259–1272, 2012. [Online]. Available:

[81] I. Rabinowitch, M. Chatzigeorgiou, B. Zhao, M. Treinin, and W. R. Schafer, “Rewiring neural circuits by the insertion of ectopic electrical synapses in transgenic C. elegans,” Nature Commun., vol. 5, no. 1, 2014, Art. no. 4442. [Online]. Available: details/publication/pub.1050134945 and PMC4109004

[82] M. Z. Miskin et al., “Electronically integrated, mass-manufactured, microscopic robots,” Nature, vol. 584, no. 7822, pp. 557–561, Aug. 2020.

[83] N. Kumar, H. Soni, S. Ramaswamy, and A. K. Sood, “Flocking at a distance in active granular matter,” Nature Commun., vol. 5, no. 1, p. 4688, Dec. 2014.

[84] M. Bär, R. Großmann, S. Heidenreich, and F. Peruani, “Self-propelled rods: Insights and perspectives for active matter,” Annu. Rev. Condens. Matter Phys., vol. 11, no. 1, pp. 441–466, Mar. 2020.

[85] R. Jeanson and A. Weidenmüller, “Interindividual variability in social insects–proximate causes and ultimate consequences,” Biol. Rev., vol. 89, no. 3, pp. 671–687, 2014. [Online]. Available: https://

[86] I. D. Chase, “Social process and hierarchy formation in small groups: A comparative perspective,” Amer. Sociol. Rev., vol. 45, no. 6, pp. 905–924, 1980. [Online]. Available:

[87] G. Theraulaz, E. Bonabeau, and J.-L. Deneubourg, “Self-organization of hierarchies in animal societies: The case of the primitively eusocial wasp Polistes dominulus Christ,” J. Theor. Biol., vol. 174, no. 3, pp. 313–323, 1995. [Online]. Available: science/article/pii/S0022519385701014

[88] N. Mathews, A. L. Christensen, R. O’Grady, F. Mondada, and M. Dorigo, “Mergeable nervous systems for robots,” Nature Commun., vol. 8, no. 1, p. 439, Dec. 2017.

[89] W. Zhu, M. Allwright, M. K. Heinrich, S. O˘guz, A. L. Christensen, and M. Dorigo, “Formation control of UAVs and mobile robots using self-organized communication topologies,” in Proc. 12th Int. Conf. Swarm Intell. (ANTS) (Lecture Notes in Computer Science), vol. 12421. Berlin, Germany: Springer, 2020, pp. 306–314.

[90] A. Attanasi et al., “Finite-size scaling as a way to probe near-criticality in natural swarms,” Phys. Rev. Lett., vol. 113, Dec. 2014, Art. no. 238102, doi: 10.1103/PhysRevLett.113.238102.

[91] M. A. Muñoz, “Colloquium: Criticality and dynamical scaling in living systems,” Rev. Mod. Phys., vol. 90, Jul. 2018, Art. no. 031001, doi: 10.1103/RevModPhys.90.031001.

[92] D. S. Calovi, U. Lopez, P. Schuhmacher, H. Chaté, C. Sire, and G. Theraulaz, “Collective response to perturbations in a data-driven fish school model,” J. Roy. Soc. Interface, vol. 12, no. 104, 2015, Art. no. 20141362. [Online]. Available: 10.1098/rsif.2014.1362

[93] L. Busoniu, R. Babuska, and B. De Schutter, “A comprehensive survey of multiagent reinforcement learning,” IEEE Trans. Syst., Man, Cybern. C, Appl. Rev., vol. 38, no. 2, pp. 156–172, Mar. 2008.

[94] E. Price et al., “Deep neural network-based cooperative visual tracking through multiple micro aerial vehicles,” IEEE Robot. Autom. Lett., vol. 3, no. 4, pp. 3193–3200, Oct. 2018.

[95] F. Magistri, D. Nardi, and V. Trianni, “Using prior information to improve crop/weed classification by MAV swarms,” in Proc. 11th Int. Micro Air Vehicle Competition Conf., Madrid, Spain, 2019, pp. 67–75.

[96] E. R. Hunt and S. Hauert, “A checklist for safe robot swarms,” Nature Mach. Intell., vol. 2, no. 8, pp. 420–422, Aug. 2020.

[97] F. Higgins, A. Tomlinson, and K. M. Martin, “Survey on security challenges for swarm robotics,” in Proc. 5th Int. Conf. Auton. Auton. Syst., 2009, pp. 307–312.

[98] E. C. Ferrer, T. Hardjono, M. Dorigo, and A. S. Pentland, “Secure and secret cooperation in robotic swarms,” 2019, arXiv:1904.09266. [Online]. Available: 1904.09266

[99] V. Strobel, E. C. Ferrer, and M. Dorigo, “Blockchain technology secures robot swarms: A comparison of consensus protocols and their resilience to Byzantine robots,” Frontiers Robot. AI, vol. 7, p. 54, May 2020.

[100] A. Prorok and V. Kumar, “Towards differentially private aggregation of heterogeneous robots,” in Proc. 13th Int. Symp. Distrib. Auton. Robotic Syst., R. Groß, A. Kolling, S. Berman, E. Frazzoli, A. Martinoli, F. Matsuno, and M. Gauci, Eds. Cham, Switzerland: Springer, 2018, pp. 587–601, doi: 10.1007/978-3-319-73008-0_41

[101] V. Strobel, E. C. Ferrer, and M. Dorigo, “Managing Byzantine robots via blockchain technology in a swarm robotics collective decision making scenario,” in Proc. 17th Int. Conf. Auton. Agents Multiagent Syst. (AAMAS), M. Dastani, G. Sukthankar, E. André, and S. Koenig, Eds. Richland, SC, USA: International Foundation for Autonomous Agents and Multiagent Systems, 2018, pp. 541–549.

[102] C. Bartneck, T. Belpaeme, F. Eyssel, T. Kanda, M. Keijsers, and S. Šabanovi´c, Human-Robot Interaction: An Introduction. Cambridge, U.K.: Cambridge Univ. Press, 2020.

[103] D. S. Brown, M. A. Goodrich, S.-Y. Jung, and S. C. Kerman, “Two invariants of human swarm interaction,” J. Human-Robot Interact., vol. 5, no. 1, pp. 1–31, Aug. 2015.

[104] A. Kolling, P. Walker, N. Chakraborty, K. Sycara, and M. Lewis, “Human interaction with robot swarms: A survey,” IEEE Trans. Human-Mach. Syst., vol. 46, no. 1, pp. 9–26, Feb. 2016.

[105] J. Gautrais, C. Jost, R. Jeanson, and G. Theraulaz, “How individual interactions control aggregation patterns in gregarious arthropods,” Interact. Stud., vol. 5, no. 2, pp. 245–269, 2004. [Online]. Available: content/journals/10.1075/is.5.2.05gau

[106] I. D. Couzin, J. Krause, N. R. Franks, and S. A. Levin, “Effective leadership and decision-making in animal groups on the move,” Nature, vol. 433, no. 7025, pp. 513–516, Feb. 2005.

[107] A. Baronchelli, “The emergence of consensus: A primer,” Roy. Soc. Open Sci., vol. 5, no. 2, Feb. 2018, Art. no. 172189.

[108] G. Podevijn, R. O’Grady, Y. S. G. Nashed, and M. Dorigo, “Gesturing at subswarms: Towards direct human control of robot swarms,” in Proc. 14th Annu. Conf. Towards Auton. Robotic Syst. (TAROS) (Lecture Notes in Computer Science), vol. 8069, A. Natraj, S. Cameron, C. Melhuish, and M. Witkowski, Eds., 2013, pp. 390–403.

[109] J. Nagi, A. Giusti, L. M. Gambardella, and G. A. Di Caro, “Human-swarm interaction using spatial gestures,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst., Sep. 2014, pp. 3834–3841.

[110] L. Mondada, M. E. Karim, and F. Mondada, “Electroencephalography as implicit communication channel for proximal interaction between humans and robot swarms,” Swarm Intell., vol. 10, no. 4, pp. 247–265, Dec. 2016.

[111] G. Podevijn, R. O’Grady, N. Mathews, A. Gilles, C. Fantini-Hauwel, and M. Dorigo, “Investigating the effect of increasing robot group sizes on the human psychophysiological state in the context of human–swarm interaction,” Swarm Intell., vol. 10, no. 3, pp. 193–210, Sep. 2016.

[112] G. Podevijn, R. O’Grady, C. Fantini-Hauwel, and M. Dorigo, “Human responses to stimuli produced by robot swarms—The effect of the reality-gap on psychological state,” in Proc. 13th Int. Symp. Distrib. Auto. Robotic Syst. (Springer Proceedings in Advanced Robotics), vol. 6, R. Groß, A. Kolling, S. Berman, E. Frazzoli, A. Martinoli, F. Matsuno, and M. Gauci, Eds. Cham, Switzerland: Springer, 2018, pp. 531–543.

[113] C. Nam, P. Walker, H. Li, M. Lewis, and K. Sycara, “Models of trust in human control of swarms with varied levels of autonomy,” IEEE Trans. HumanMach. Syst., vol. 50, no. 3, pp. 194–204,Jun.2020.

[114] M. Dorigo, G. Theraulaz, and V. Trianni, “Reflections on the future of swarm robotics,” Sci. Robot., vol. 5, no. 49, Dec. 2020, Art. no. eabe4385.

[115] M. Dorigo and E. Sahin, Eds., “Special issue on swarm robotics,” Auton. Robots, vol. 17, nos. 2–3, pp. 111–246, Sep. 2004.

[116] G.-Z. Yang et al., “The grand challenges of science robotics,” Sci. Robot., vol. 3, no. 14, 2018, Art. no. eaar7650. [Online]. Available: https://

[117] S. Kazadi, “Swarm engineering,” Ph.D. dissertation, California Inst. Technol., Pasadena, CA, USA, 2000.

[118] M. Li et al., “Generating swarm solution classes using the Hamiltonian method of swarm design,” in Proc. 9th Int. Conf. Agents Artif. Intell., vol. 2. Setúbal, Portugal: SciTePress, 2017, pp. 145–152.

[119] A. Renzaglia, L. Doitsidis, A. Martinelli, and E. B. Kosmatopoulos, “Multi-robot three-dimensional coverage of unknown areas,” Int. J. Robot. Res., vol. 31, no. 6, pp. 738–752, May 2012.

[120] W. Hönig, J. A. Preiss, T. K. S. Kumar, G. S. Sukhatme, and N. Ayanian, “Trajectory planning for quadrotor swarms,” IEEE Trans. Robot., vol. 34, no. 4, pp. 856–869, Aug. 2018.

[121] W. Hönig, S. Kiesel, A. Tinka, J. W. Durham, and N. Ayanian, “Persistent and robust execution of MAPF schedules in warehouses,” IEEE Robot. Autom. Lett., vol. 4, no. 2, pp. 1125–1131, Apr. 2019.

[122] R. R. Negenborn and J. M. Maestre, “Distributed model predictive control: An overview and roadmap of future research opportunities,” IEEE Control Syst., vol. 34, no. 4, pp. 87–97, Aug. 2014.

[123] C. E. Luis, M. Vukosavljev, and A. P. Schoellig, “Online trajectory generation with distributed model predictive control for multi-robot motion planning,” IEEE Robot. Autom. Lett., vol. 5, no. 2, pp. 604–611, Apr. 2020.

[124] H. Hamann and A. Reina, “Scalability in computing and robotics,” 2020, arXiv:2006.04969. [Online]. Available:

[125] D. Albani, T. Manoni, D. Nardi, and V. Trianni, “Dynamic UAV swarm deployment for non-uniform coverage,” in Proc. 17th Int. Conf. Auton. Agents MultiAgent Syst. (AAMAS), 2018, pp. 523–531.

[126] N. Correll and A. Martinoli, “Multirobot inspection of industrial machinery,” IEEE Robot. Autom. Mag., vol. 16, no. 1, pp. 103–112, Mar. 2009.

[127] J. J. Dawkins, F. L. Crabbe, and D. Evangelista, “Deployment and flight operations of a large scale UAS combat swarm: Results from DARPA service academies swarm challenge,” in Proc. Int. Conf. Unmanned Aircr. Syst. (ICUAS), Jun. 2018, pp. 1271–1278.

[128] G. C. Messenger, “Radiation hardening,” in AccessScience. McGraw-Hill, 2020. [Online]. Available: content/566850

[129] T. M. Roehr, F. Cordes, and F. Kirchner, “Reconfigurable integrated multirobot exploration system (RIMRES): Heterogeneous modular reconfigurable robots for space exploration,” J. Field Robot., vol. 31, no. 1, pp. 3–34, 2014. [Online]. Available: https://onlinelibrary.wiley. com/doi/abs/10.1002/rob.21477

[130] C. Pinciroli et al., “Self-organizing and scalable shape formation for a swarm of pico satellites,” in Proc. NASA/ESA Conf. Adapt. Hardw. Syst., Los Alamitos, CA, USA: IEEE Computer Society Press, Jun. 2008, pp. 57–61.

[131] M. Waibel, B. Keays, and F. Augugliaro, “Drones shows: Creative potential and best practices,” Verity Studios, Zurich, Switzerland, White Paper, Jan. 2017.

[132] J. Alonso-Mora, A. Breitenmoser, M. Rufli, R. Siegwart, and P. Beardsley, “Image and animation display with multiple mobile robots,” Int. J. Robot. Res., vol. 31, no. 6, pp. 753–773, May 2012.

[133] M. Le Goc, L. H. Kim, A. Parsaei, J.-D. Fekete, P. Dragicevic, and S. Follmer, “Zooids: Building blocks for swarm user interfaces,” in Proc. 29th Annu. Symp. User Interface Softw. Technol. New York, NY, USA: Association for Computing Machinery, Oct. 2016, p. 97.

[134] M. Alhafnawi, S. Hauert, and P. O’Dowd, “Robotic canvas: Interactive painting onto robot swarms,” in Proc. Conf. Artif. Life, 2020, pp. 163–170.

[135] E. Tsykunov, R. Agishev, R. Ibrahimov, L. Labazanova, A. Tleugazy, and D. Tsetserukou, “SwarmTouch: Guiding a swarm of micro-quadrotors with impedance control using a wearable tactile interface,” IEEE Trans. Haptics, vol. 12, no. 3, pp. 363–374, Jul. 2019.

[136] W. Hönig, C. Milanes, L. Scaria, T. Phan, M. Bolas, and N. Ayanian, “Mixed reality for robotics,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst. (IROS), Sep. 2015, pp. 5382–5387.

[137] S. Jeon et al., “Magnetically actuated microrobots as a platform for stem cell transplantation,” Sci. Robot., vol. 4, no. 30, May 2019, Art. no. eaav4317.

[138] X. Dong and M. Sitti, “Controlling two-dimensional collective formation and cooperative behavior of magnetic microrobot swarms,” Int. J. Robot. Res., vol. 39, no. 5, 2020, Art. no. 027836492090310.

Essay topic:

Swarm Robotics: Past, Present, and Future

Paper address:

Posted by:CoinYuppie,Reprinted with attribution to:
Coinyuppie is an open information publishing platform, all information provided is not related to the views and positions of coinyuppie, and does not constitute any investment and financial advice. Users are expected to carefully screen and prevent risks.

Like (0)
Donate Buy me a coffee Buy me a coffee
Previous 2021-07-22 06:09
Next 2021-07-22 06:13

Related articles