“…The ‘Machine learning’ method can be used for aggregation (Trianni et al, 2003), flocking (Baldassarre et al, 2003; Salimi & Pasquier, 2021) and pattern formation (Sharma et al, 2022). It is reasonable to assume that the other behaviours can be obtained with the right fitness function or the correct reward mechanism.…”
Section: Discussionmentioning
confidence: 99%
“…Deep learning is used to help the robot find the best policy by representing this policy with a neural network of which the weights are adjusted through time. Flocking (Salimi & Pasquier, 2021) and pattern formation (Sharma et al, 2022) are two examples of collective behaviours that can be obtained with this method and process.…”
Self-organisation in robot swarms can produce collective behaviours, particularly through spatial self-organisation. For example, it can be used to ensure that the robots in a swarm move collectively. However, from a designer’s point of view, understanding precisely what happens in a swarm that allows these behaviours to emerge at the macroscopic level remains a difficult task. The same behaviour can come from multiple different controllers (ie the control algorithm of a robot) and a single controller can give rise to multiple different behaviours, sometimes caused by slight changes in self-organisation. To grasp the causes of these differences, it is necessary to investigate the relationships between the many methods of self-organisation that exist and the various behaviours that can be obtained. The work presented here addresses self-organisation in robot swarms by focusing on the main behaviours that lead to spatial self-organisation of the robots. First, we propose a unified definition of the different behaviours and present an original classification system highlighting ten self-organisation methods that each allow one or more behaviours to be performed. An analysis, based on this classification system, links the identified mechanisms with behaviours that could be considered as obtainable or not. Finally, we discuss some perspectives on this work, notably from the point of view of an operator or designer.
“…The ‘Machine learning’ method can be used for aggregation (Trianni et al, 2003), flocking (Baldassarre et al, 2003; Salimi & Pasquier, 2021) and pattern formation (Sharma et al, 2022). It is reasonable to assume that the other behaviours can be obtained with the right fitness function or the correct reward mechanism.…”
Section: Discussionmentioning
confidence: 99%
“…Deep learning is used to help the robot find the best policy by representing this policy with a neural network of which the weights are adjusted through time. Flocking (Salimi & Pasquier, 2021) and pattern formation (Sharma et al, 2022) are two examples of collective behaviours that can be obtained with this method and process.…”
Self-organisation in robot swarms can produce collective behaviours, particularly through spatial self-organisation. For example, it can be used to ensure that the robots in a swarm move collectively. However, from a designer’s point of view, understanding precisely what happens in a swarm that allows these behaviours to emerge at the macroscopic level remains a difficult task. The same behaviour can come from multiple different controllers (ie the control algorithm of a robot) and a single controller can give rise to multiple different behaviours, sometimes caused by slight changes in self-organisation. To grasp the causes of these differences, it is necessary to investigate the relationships between the many methods of self-organisation that exist and the various behaviours that can be obtained. The work presented here addresses self-organisation in robot swarms by focusing on the main behaviours that lead to spatial self-organisation of the robots. First, we propose a unified definition of the different behaviours and present an original classification system highlighting ten self-organisation methods that each allow one or more behaviours to be performed. An analysis, based on this classification system, links the identified mechanisms with behaviours that could be considered as obtainable or not. Finally, we discuss some perspectives on this work, notably from the point of view of an operator or designer.
“…Results from three robots show that the proposed algorithm improves over the standard MADDPG. Similarly, Salimi and Pasquier [ 106 ] have proposed the use of DDPG with centralized training and a decentralized execution mechanism to train the flocking policy for a system of UAVs. Such flocking with UAVs might be challenging due to complex kinematics.…”
Section: Multi-robot System Applications Of Multi-agent Deep Reinforc...mentioning
Deep reinforcement learning has produced many success stories in recent years. Some example fields in which these successes have taken place include mathematics, games, health care, and robotics. In this paper, we are especially interested in multi-agent deep reinforcement learning, where multiple agents present in the environment not only learn from their own experiences but also from each other and its applications in multi-robot systems. In many real-world scenarios, one robot might not be enough to complete the given task on its own, and, therefore, we might need to deploy multiple robots who work together towards a common global objective of finishing the task. Although multi-agent deep reinforcement learning and its applications in multi-robot systems are of tremendous significance from theoretical and applied standpoints, the latest survey in this domain dates to 2004 albeit for traditional learning applications as deep reinforcement learning was not invented. We classify the reviewed papers in our survey primarily based on their multi-robot applications. Our survey also discusses a few challenges that the current research in this domain faces and provides a potential list of future applications involving multi-robot systems that can benefit from advances in multi-agent deep reinforcement learning.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.