LEAF: A Cloud-end Coevolutionary Framework for Multi-objective Swarm Robots Formation Optimization in Dynamic Environment

Swarm robots systems are of great value in many fields such as transportation, rescue, and military. In its application scenarios, formation is an indispensable ability. Nowadays, rapidly developing cloud computing brings new vigor to swarm robots formation. However, simply superimposing cloud computing and swarm robots together cannot solve the challenges of untimely cloud response, multiple formation objectives, and un-predictable environmental changes. Therefore, this paper focuses on the problem of swarm robots formation and proposes a set of feasible solutions. In particular, based on the concept of cloud-end synchronization, a decision-control decoupling cloud-end coevolutionary framework, LEAF, is proposed. This framework can make full use of cloud resources to make decisions while allowing the robots to make autonomous control based on their observed environment, hence flexibly responding t 0 the dynamic environment. On this basis, we design a corresponding algorithm, LEAF-GV. In LEAF-GV, NSGA-II is used for cloud decision-making, and the virtual potential field is used for robots autonomous controlling so that the swarm robots can take into account different formation objectives in a dynamic environment. According to our experimental results, the proposed framework and method can effectively adapt to the formation needs of swarm robots.