【作者】 刘国良； 孙增圻；

【Author】 Liu Guoliang,Sun Zengqi(State Key Laboratory of Intelligent Technology and Systems,Graduate School at Shenzhen,Tsinghua University,Shenzhen 518055,China)

【机构】 清华大学深圳研究生院智能技术与系统国家重点实验室；

【摘要】 移动机器人利用超声波传感器在狭窄空间中导航,超声波传感器的镜面反射问题非常严重,使移动机器人不能正确地认识周围的环境。针对已知狭窄空间,本文在移动机器人进行导航之前使用超声波传感器来建立一个准确的环境地形图,然后让移动机器人记住这个地形图。这个地形图在移动机器人完成规定任务时,例如路径规划、避障等,能够提供可靠的周围环境信息,不需要机器人随时检测周围环境的情况,从而节省了处理器的时间,并避免了现场局部检测的不确定性问题。在建立地形图过程中使用了一种多层融合算法,该算法通过在数据层融合中使用D-S证据理论,在决策层融合中使用意见一致性理论,有效地减少了由于超声波传感器镜面反射所引起的不确定性。最后,实验证明了新方法的有效性。更多还原

【Abstract】 When mobile robots navigate in a confined space, the problem arising from specular reflection of ultrasonic sensors is so serious that mobile robots can’t recognize their surrounding environment correctly. If the environment is known, we build its map with ultrasonic sensors in advance before navigation is performed, and then let mobile robot memorize this map. This map can provide reliable information about the environment and mobile robots don’t need to sense the environment any longer when mobile robots perform tasks. A multi-layer fusion algorithm is proposed for building maps. Uncertainties arising from specular reflection of ultrasonic sensor are successfully reduced by using Dempster-Shafer evidence theory into data-level fusion and by using consensus theory into decision-level fusion in this algorithm. The experiment results indicated that the multi-layer fusion algorithm improved the performance of ultrasonic sensors.更多还原