论文部分内容阅读
提出一种迷宫机器人的人工脑系统,包括迷宫路标感知单元与行为决策单元。其中,感知单元基于ART1神经网络,用于识别迷宫导向路标;决策单元基于行为概率实现矩阵,并以强化学习更新行动策略。机器人所在迷宫的特征为每个路口设有导向路标,路标为含噪声的符号图像。在仿真实验中,令机器人在迷宫中随机行走,通过调节人工脑系统的试验参数,经过一段时间的自主探索学习过程机器人能最终穿越迷宫。仿真实验结果表明,该人工脑系统能够自组织地理解迷宫中导向路标的含义,并引导机器人成功穿越迷宫。同时,该人工脑系统对于基于路标导航的城市巡逻机器人、高危复杂环境下的抢险机器人的研究发展有一定的推动作用。
An artificial brain system of maze robot is proposed, which includes the Labyrinth Road Marking Unit and Behavior Decision Unit. Among them, the sensing unit is based on the ART1 neural network for identifying the maze guidepost; the decision unit realizes the matrix based on the behavior probability and updates the action strategy with reinforcement learning. The maze where the robot is located features a guidepost for each intersection, which is a noisy symbolic image. In the simulation experiments, the robots are allowed to walk randomly in the maze. By adjusting the experimental parameters of the artificial brain system, the robot can eventually pass through the labyrinth after a period of autonomous exploration and learning. Simulation results show that the artificial brain system can self-organize the understanding of the meaning of the guidepost in the maze and guide the robot to successfully cross the maze. At the same time, the artificial brain system can promote the research and development of city patrol robots based on road sign navigation and rescue robots in high-risk and complex environment.