基于未知环境碰撞冲突预测的群机器人多目标搜索研究

群机器人在未知动态环境下进行多目标搜索时，存在碰撞预测和搜索效率不高等问题。提出了一种碰撞几何锥和改进惯性权重的粒子群优化算法相结合的多目标搜索策略。首先，根据静、动态威胁物的不同分别引入碰撞锥(CC)和速度障碍法(VO),提出了简化复杂障碍物的膨胀几何法(SG)和一种改进CC和VO的碰撞几何锥模型(CGC);有效解决了复杂不规则威胁物的避碰预测问题，并根据CGC模型作出威胁评估报告以确定最优避障方向。其次，提出一种改进惯性权重的粒子群优化算法(IWPSO),提高了搜索效率同时有效解决了粒子群优化算法易陷入局部最优的问题。最后，将两种改进的方法(CGC-IWPSO)相结合以实现群机器人的多目标任务搜索，相比于简化虚拟受力(SVF)、自适应机器人蝙蝠算法(ARBR)、具有运动学约束的粒子群算法(KCPSO),本文方法在搜索时耗、能耗以及避障次数上分别至少减少了15.59%、 10.14%、 14.12%。

Research on Multi-objective Search of Swarm Robots Based on Collision Conflict Prediction in Unknown Environment

The application of swarm robots in multi-objective search is an unexplored dynamic environment with several issues, such as collision prediction and low search efficiency. We proposes a multi-objective search strategy by combining the collision geometric cone and particle swarm optimization algorithm with improved inertia weight. First, the collision cone (CC) and the velocity obstacle method (VO) are introduced using the difference between static and dynamic threats. Furthermore, the swelling geometry method is proposed to simplify complex obstacles, and a collision geometric cone model (CGC) is applied to improve CC and VO. This model can effectively resolve the issue of predicting collision avoidance of complex irregular threats and can determine the optimal obstacle avoidance direction from the threat assessment report based on the CGC model.