Group Self-organized Confrontation In Multiple Mobile Robots System

Multiple mobile robots system is a distributed system that can carry out the complex cooperation tasks with multiple agents.Each agent has simple functions and local communication.The system can complete complicated tasks which are beyond the capacity of a single agent.This dissertation addresses the group selfconfrontation in multiple mobile robots system.An in-depth study on three typical scenarios,including pursuit-evasion game,defense-intrusion game,and entrapment game are provided to investigate the swarm intelligence in group self-confrontation.The main contributions and research achievements of this dissertation are as follows:(1)Escape strategy and anti-attack behavior in a pursuit-evasion gameImmediate flight upon detection is a common but not exclusive response for prey in pursuit-evasion game with multiple mobile robots.Two escape strategies: stayeat strategy and run-away strategy are proposed for escape robots with regards to energy management.By investigating the survivability of the escape robot group under different escape strategies with different parameters,the critical conditions of choosing a better escape strategy are obtained.The results show that the optimal escape speed of the escape robot is context-dependent rather than the fastest always producing the best chance of survival.The stay-eat or run-away decision depends mainly on the maximum speed and the energy dissipation ratio of prey to predator.Stay-eat behavior is more effective when the prey have a higher escape speed and a higher energy dissipation rate,where aggregation can induce this stay-eat behavior.Moreover,an anti-attack pursuit-evasion game model is structured to investigate the effect of anti-attack behavior on the dynamics of the pursuit-evasion game.The results show that anti-attack behavior contributes to enhance the survivability of the escape robot group and the effect is more efficient in the presence of aggregation preference.In this point,more reasonable aggregation strategy is designed for escape robot group.(2)Motion patterns and synchronous intercept strategy in a defense-intrusion gameWe propose and formulate a defense-intrusion game with synchronous intercept constraint,where a single intruder tries to reach a stationary target that are protected by a swarm of defenders,and the defenders try to capture the intruder by a synchronous attack(at least n defenders have to synchronously reach the intruder’s location to capture it).Three general collective motions with simple attraction-repulsion interaction between defender pairs are introduced to investigate the dynamics and motion patterns of the defense-intrusion game.In addition,two kinds of synchronous intercept strategies are proposed for the defense-intrusion game,introduced here as attacker-oriented strategy and neutral-position-oriented strategy.Theoretical analysis and simulation results show that the two strategies are able to generate different synchronous intercept patterns: contact intercept pattern and stable non-contact intercept pattern,respectively.The boundary condition,the efficiency and the maximum tolerable noise of synchronous intercept strategies are also obtained.(3)Multi-target trapping with swarm robots in complex environmentA significant limitation of most conventional target trapping algorithms for swarm robots is that the target shape needs to be predefined and only some regular Euclidean shapes can be applied.This may be inadequate for dealing with the problem of entrapment in dynamic targets.We proposes a flexible shape formation algorithm by using Radial Basis Implicit Function to realize the multi-target trapping task that needs the transformation of trapping shape response to dynamic targets.With this flexible shape formation method,most distribution of group targets are able to be entrapped,also splitting and merging of multiple shapes depending on moving targets can be achieved.Numerical simulations of static/dynamic scenarios,obstacle avoidance,noise and self-reorganization have been performed to validate the effectiveness and flexibility of the proposed approach for multi-target trapping.