摘要
为了提高自主移动机器人的动态避障能力,提出了一种基于视觉信息的拟人动态避障方法.以视觉信息为决策依据,利用神经网络对障碍物的屏幕坐标和实际相对坐标进行非线性映射,在避障过程中只考虑障碍物相对于机器人的运动及可能的碰撞方式,无需考虑机器人和障碍物的运动速度和运动方向.仿真试验证明这种方法是可行而有效的.
In order to improve the capability of dynamic obstacle avoidance for autonomous mobile robots, a method of dynamic obstacle avoidance that simulates human's behavior is presented. This method makes decisions according to the vision information acquired and realizes nonlinear mapping from screen coordination to actual corresponding coordination. During the course of the obstacle avoidance, this method makes the path planning more quick, because it only calculates the obstacles' corresponding velocity, direction and probable collision modes, without taking into account the velocity and direction of the robot and the obstacles. The experiment of simulation indicates that the method is feasible and valid.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
2004年第7期854-856,共3页
Journal of Harbin Institute of Technology
基金
国家高技术研究发展计划资助项目(863-2001AA422270).
关键词
自主移动机器人
人工神经网络
动态避障
Backpropagation
Collision avoidance
Computer simulation
Motion planning
Neural networks
