Research And Implementation Of Key Technologies Of Ground Mobile Robot Navigation In Indoor Environments

With the rapid development of robotics,service robots have gradually taken part in people's daily life,performing tasks such as cleaning,care,and security.Compared with industrial robots,working environments of service robots are usually unpredictable,unstructured and dynamic.Meanwhile,in order to complete corresponding tasks,ser-vice robots also require autonomous mobility.Consequently,environment reconstruction and autonomous navigation are fundamental technologies of service robots.On the other hand,robot sensors able to acquire three-dimensional information get popularized due to their reduced costs.Therefore,making full use of three-dimensional sensors and extend-ing existing two-dimensional robot navigation techniques into three-dimensional space are of theoretical and practical importance for service robot applications.By taking advantages of the relatively simpler working environment,an eff-icient three-dimensional space representation called multi-level occupancy grids(M-LOG)is proposed for indoor ground mobile robots.Based on two-dimensional occupancy grids commonly used in mapping and path planning,M-LOG stores environmental information from different heights in an perfectly aligned stack of planar slices,and can effectively and efficiently represent the key information in three-dimensional environments.A simultaneous localization and mapping(SLAM)approach based on M-LOG has been designed and implemented to meet the requirements of indoor ground mobile robot-s.This approach can not only quickly reconstruct a map representations containing im-portant 3D information of the environment,but also improve the accuracy of the robot localization and mapping results by using multi-level scan data.A trafficability-aware navigation method based on M-LOG has been designed and implemented.In this method,traff-icability affected by robot characteristics including geometric shapes and obstacle capabilities are introduced into the system as constraints.M-LOG can be transferred into a trafficability-aware costmap,upon which robot-specific navigation can be performed.Finally,a Turtlebot 2 robot running the open-source robot operating system(ROS)is used as the experimental platform,and experiments on SLAM and autonomous nav-igation based on M-LOG are carried out in both the Gazebo simulator and real world.Experimental and testing results suggest that M-LOG is able to provide an efficient yet effective representation for mobile robots working in indoor environments.It also helps in improving the SLAM accuracy and planning optimality.Compared with other typical 3D representations,M-LOG,as well as related mapping and navigation algorithms scale linearly in both storage and computation with key parameters,including level counts and particle numbers.