School of Electrical Engineering, Zhengzhou University
针对多机器人在未知区域的覆盖搜索问题,提出一种基于生物启发神经网络和分布式模型预测控制(DMPC)的多机器人协同搜索算法. 首先用栅格地图表示未知区域,然后基于栅格地图建立起生物启发神经网络来表示动态搜索环境. 在生物启发神经网络中,未搜索栅格的神经元活性值大于已搜索栅格和障碍物栅格. 在此基础上,为了平衡机器人覆盖搜索过程中的短期收益和长期收益,避免后期陷入局部最优,引入DMPC作为决策方法. 选择预测周期内机器人所覆盖栅格的神经元活性值增量作为主要激励函数,引导机器人向未覆盖区域搜索. 通过采用差分进化算法(DE)进行优化求解,得到最优解. 最后通过设计仿真实验,验证了该方法的有效性和优越性.
To solve the problem of multi-robot coverage search in unknown areas, a multi-robot cooperative search algorithm based on bio-inspired neural network and distributed model predictive control (DMPC) is proposed. Firstly, the unknown region is represented by raster map, and then the bio-inspired neural network is established based on raster map to represent dynamic search environment. In the bio-inspired neural network, the activity value of unsearched grids is higher than searched grids and obstacle grids. On this basis, in order to balance the short-term gains and long-term gains in the process of robot coverage search,and avoid falling into local optimization in the later period, DMPC is introduced as the decision-making method. The increment of neuron activity value of the raster covered by the robot in the forecast period is selected as the main excitation function to guide the robot to search the uncovered area. The optimal solution is obtained by using differential evolutionary algorithm(DE). Finally, the effectiveness and superiority of the proposed method are verified by designing simulation experiments.