Multi-Agent Pursuit-Evasion Problem Considering The Uncertainty Of Pursuers' Capabilities

Multiagent pursuit-evasion problem is a classical problem in the fields of artificial intelligence and distributed collaboration.This kind of problems is related to many real-world pursuit-evasion processes(for instance,the police pursue the offender).In real world,if single pursuer cannot capture the evader,multiple pursuers need to collaboratively capture the evader.Therefore,many previous studies have focused on designing effective collaborative strategy for the pursuers.When designing collaborative pursuing strategy,the capabilities of the pursuers should be considered.It is worth noting that,previous studies usually assume that the the capabilities of the pursuers are deterministic.However,in some real situations,this assumption is invalid,i.e.,the capabilities of the pursuers may be uncertain.Because previous pursuing strategies usually depend on the pursuers' deterministic capacities,the performance of these strategies will be influenced by the uncertainty of the pursuers' capacities.Therefore,this paper studies the multiagent pursuit-evasion problem with the uncertainty of the pursuers' capacities.Based on related studies,we model the uncertainty of the pursuers' capacities and present feasible pursuing strategies.The detailed contributions are as follows:· The multiagent pursuit-evasion problem with the uncertainty of the pursuers' speeds is studied.We design the strategy that can address the uncertainty of pursuers' speeds.Experiments show that the proposed strategy can outperform the benchmark strategies when the pursuer's speeds are uncertain.· The multiagent pursuit-evasion problem with energy-limited pursuers and uncertainty of energy consumption is studied.We systematically consider different situations and design the sub-algorithms including speed-control algorithm and path planning algorithm.Then,the sub-algorithms are integrated into a self-adaptable cooperation strategy based on simulated annealing.Experiments show that the proposed strategy can outperform the benchmark strategies and be better adaptive to the variety of the pursuers' maximum speed and the variety of the escape strategies.· The multiagent pursuit-evasion problem with the uncertainty of the pursuers' contending ability is studied.We systematically analyze different subcases of this problem and then design the pursuing strategies respectively feasible in these subcases.Experiments show that the proposed strategy can outperform the benchmark strategies.Especially,when the environment significantly influences the results of the conflicts between the pursuers and the evader,the advantage of the proposed strategies is more obvious.· The multiagent pursuit-evasion problem with the uncertainty of the pursuers' sneak ability is studied.We design the strategies respectly based on two typical ideas.Based on the experiments,we analyzed the situations where the two strategies respectively applicative.Overall,we systematically analyze the influence of the uncertainties of the pursuers' capacities.We design the pursuing strategies which can address the uncertainties of the pursuers' capacities.Our work can facilitate the theoretical researches of pursuit-evasion problem and the applications of the pursuer robot teams.