Behavior Learning of Multiagent Robots Based on Sensor Fusion of Environment Information

针对移动机器人的行为学习问题，研究了有限方向运动的遗传路径规划算法和基于传感器融合的跟踪控制方案。该控制方案简单有效，可以实时应用，并使传感器融合网络学习规划路径、机器人运动和传感器信号之间的关系。在黑条纹标记上的直线和圆形运动用于教学网络，其中内部传感器与外部传感器融合。初步实验解释了机器人运动与传感器输出之间的关系。实验结果表明，采用红外光反射器和陀螺传感器的测量系统是有效的，同时也验证了传感器融合网络学习方法的可行性。提出了一种基于单片机网络的模块化控制系统，并利用串行总线对其性能进行了实验评价。提出了一种空分红外光通信系统，作为机器人间局部通信的一种方法。通信系统使用一组模块或收发器，这些模块或收发器安装在机器人身体的圆周上。每个模块检测红外线的入射角。由于一个机器人可以同时通过不同的模块与多个机器人进行通信，因此可以创建空间划分通信网络。硬件实现和性能测量表明，1)机器人之间的通信连接在移动时保持不中断，2)可以以±5°的精度检测其他机器人的方向。

As a study of behavior learning of mobile robots, a genetic path planning algorithm with limited directional movements and a sensor fusion based tracking control scheme were examined. The control scheme is simple and efficient enough to be used in real time, and it makes the sensor fusion network learn relations between the planned path, the robot movement, and the sensor signals. Straight-line and circular movements on black-striped markers were used for teaching the network, where the internal sensors are fused with the external sensors. The preliminary experiments explain relations between the robot movement and the sensor outputs. The experimental results show the effectiveness of the proposed measurement system using the infrared photo reflectors and the gyro sensor, and also the feasibility of the learning method for the sensor fusion network. A microcontroller network based modular control system was proposed, and its performance using a serial bus was experimentally evaluated. A space-division infrared optical communication system was proposed as a method of local inter-robot communication. The communication system uses a set of modules, or transceivers, which is arranged in the circumference of the robot body. Each module detects the angle of incidence of infrared rays. Space-division communication networks can be created, because a robot can communicate with more than one robot through different modules at the same time. Hardware realization and performance measurements show that 1) communication connections between robots are maintained without interruptions when they are moving, 2) the direction of other robots can be detected with an accuracy of ±5°.