Enviromental Perception Based On 3D LIDAR For Rescue Robots

Urban search and rescue(USAR)after earthshake,volcanic eruption or other disasters is at high risk due to the unstable structure of collapsed buildings.Rescue robot is playing a more important role in USAR.The ultimate objective of rescue robot is to complete the USAR tasks autonomously.3D LIDAR has advantages in higher accuracy and wider range of recognition,and therefore ensures the safty of robot better.In this paper,a series of researches on the environmental perception for rescue robots based on 3D lidar are carried out,focusing on real-time simutanious localization and mapping,realtime terrain classification and traversable area detection.Firstly,the LOAM algorithm is improved based on the fusion of IMU information for the real-time simutanious localization and mapping tasks of robots in rescue environment.The pose estimation of high frequency and low frequency map output are run on two independent threads to improve the real-time performance of the algorithm,and the IMU information fusion are used to modify the pitching motion discontinuities existing in the rescue environment.The algorithm accomplish the robot pose estimation and environment map establishment under the rescue environment in real time,and has better performance on the pitch angle motion discontinuity.Secondly,a classification method based on 3D LIDAR is proposed for the real-time terrain classification task.Combining the pose estimation and the mapping results,the region cloud of the robot is extracted.The NARF feature and spatial information are fused,and the ELM classifier is applied to classify the terrain of the robot area.Experiments show that the method has good performance in real-time classifying performance with similar classification accuracy.Thirdly,a novel method based on 3D cost map is proposed for the traversable region detection task.Based on the probability distribution of point clouds,a three-dimensional grid map is established.The cost value of each grid is calculated based on the distances between the three-dimensional grids and the nearest obstacle.Finally,the cost value of each grid is filtered to obtain the robot accessible area by a cost value threshold.The experimental results show that the algorithm is able to obtain the three-dimensional space that robots can access and meet the basic needs of rescue robots to complete the task of autonomous exploration.