无人驾驶车在越野环境中障碍身份识别  

Obstacle identification in cross-country environment for unmanned ground vehicle

作　　者：赵一兵[1] 郭烈[1] 张明恒[1] 李琳辉[1] 

机构地区：[1]大连理工大学工业装备结构分析国家重点实验室,辽宁大连116024

出　　处：《大连理工大学学报》2012年第1期132-138,共7页Journal of Dalian University of Technology

基　　金：中国航天科技集团五院总体部资助项目(WY-YY/M-200818JY001)

摘　　要：针对无人驾驶车越野条件下的环境感知问题,基于Dempster组合规则实现了障碍目标的身份识别.首先,基于CCD和激光传感器提取5个特征作为障碍物特征证据;然后,将传感器数据转换到证据空间,选用模糊插值法求取障碍物身份隶属度进而获取相关系数;再次选择经验公式,根据障碍物类型和环境加权系数计算基本概率赋值函数;最后,基于Dempster的组合规则求得融合后的总概率赋值函数,制定决策规则并识别障碍身份.实验结果表明基于D-S证据理论识别障碍物身份具有良好鲁棒性和实时性.Aiming at the problem of cross-country environment perception of unmanned ground vehicle,Dempster fusion rules are applied to identifying obstacle.Firstly,five kinds of representative features are selected based on CCD and laser sensor.Secondly,sensor data is transformed to evidence space,and the obstacle identification membership is computed by using fuzzy interpolative method,then correlation coefficient is obtained.Thirdly,according to obstacle identity and weight correlation,experimental formula is selected to compute basic probability assignment function.Finally,based on Dempster fusion rules,the ultimate basic probability assignment function is acquired,the identification and decision-making rules are set to determine obstacle classification.Test results show the good robustness and real-time property by using D-S theory to identify obstacle.

关 键 词：无人驾驶车 环境感知 D-S证据理论 基本概率赋值函数 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]