摘要
传统RBPF(Rao-Blackwellised particle filter)-SLAM(simultaneous localization and mapping)在机器人室内导航过程中因里程计读数产生的累计误差使得机器人定位和建图的精度大大降低,且传统的路径规划方法在突发环境中的动态避障效果差、实时性低.针对这一问题,提出了一种基于生物激励神经网络(biological inspired neural network,BINN)的室内实时激光SLAM控制方法,结合了RBPF-SLAM的高计算效率、高精准性与BINN的实时动态避障特性.首先在RBPF-SLAM抽取新粒子的过程中用机器人当前时刻与前一时刻的位姿差代替里程计读数作为建议分布函数的输入.然后使用BINN在动态突发环境中进行实时路径规划及避障.最后,在动态实时路径规划中利用BINN再次定位机器人,校准机器人的位姿.仿真和实际实验结果表明,基于BINN的室内实时激光SLAM控制方法的定位误差小于10%,建图误差小于9%,且规划的路径质量与传统的A;和Dijkstra算法相比,BINN算法规划的路径长度、转折次数及规划时间分别减少了7.09%、14.29%、6.97%,有效地提高了室内导航控制的定位和建图精度及动态规划的实时性.该方法同样适用于无人驾驶和物流运输等领域.
Due to the accumulation error caused by the odometer′s reading,there is a significant error in the robot′s pose estimation and mapping using the traditional Rao-Blackwellised particle filter-simultaneous localization and mapping(RBPF-SLAM)in indoor navigation;meanwhile,the conventional path planning method has low realtime performance and dynamic obstacle avoidance under sudden environmental changes.To solve this problem,we propose an indoor real-time laser SLAM control method with biological inspired neural network(BINN),which combines high computational efficiency and precision of RBPF-SLAM and BINN′s real-time dynamic obstacle avoidance.First,we use the robot′s pose difference between the current and the previous time to replace the odometer′s reading as the input of the suggested distribution function in RBPF-SLAM to extract new particles.Then,the BINN is used to avoid the dynamic obstacles in real-time path planning.Finally,the BINN is used to relocate the robot in real-time path planning and correct the robot′s pose.Comparison between simulation and real experimental results show that the robot positioning error of our proposed model is less than 10%,and the mapping error is less than 9%.Moreover,compared with A*and Dijkstra in terms of path qualities,the length,the number of turns,and the planning time of the path planned by BINN,reductions of 7.09%,14.29%,and 6.97%are obtained,respectively,which effectively improve the positioning and mapping accuracy of indoor navigation and the real-time performance of dynamic path planning.Our proposed model is also applicable in other fields,such as unmanned driving and logistics transportation.
作者
向玉云
黄铝文
XIANG Yuyun;HUANG Lüwen(College of Information Engineering,Northwest Agriculture and Forestry University,Yangling 712100,China;Key Laboratory of Agricultural Internet of Things,Ministry of Agriculture and Rural Affairs,Yangling 712100,China;Key Laboratory of Agricultural Information Perception and Intelligent Service,Yangling 712100,China)
出处
《信息与控制》
CSCD
北大核心
2021年第6期709-721,共13页
Information and Control
基金
国家自然科学基金资助项目(61903307)
中国博士后科学基金资助项目(2020M683575)。
