| Due to the complex environment and single cleaning task in the park,the traditional manual cleaning method not only occupies a lot of human and material resources,but also has limited working hours and low cleaning efficiency.The use of autonomous unmanned cleaning vehicles instead of labor can not only effectively save labor costs and improve operation efficiency,but also reduce the safety hazards caused by operation in bad weather such as high temperature and severe cold,so as to ensure the personal safety of staff.However,the complex and changeable environment of the park,the ups and downs of roads,the shelter of trees and buildings,and the passing pedestrians all bring challenges to the positioning and navigation of unmanned vehicles.The traditional positioning and navigation methods of unmanned vehicles can not meet the needs of accurate positioning in the park.Therefore,this thesis carries out in-depth research from three aspects: unmanned vehicle positioning,obstacle detection and navigation planning,and realizes an unmanned vehicle positioning and navigation system for park cleaning.The main research contents are as follows:(1)In view of the special scenes in the park,such as the ups and downs of roads,the shelter of trees and buildings,the large visual difference between day and night,and the back and forth movement of pedestrians and vehicles,clarify the tasks and indicators that the unmanned cleaning vehicle needs to complete in the park environment.Combined with the requirements of unmanned vehicle positioning and navigation tasks in the complex environment of the park,the sensor selection is determined,the sensor data communication is realized by using ROS robot operating system,and the software and hardware platform of multi-sensor fusion positioning and navigation system for the park is built.(2)Aiming at the GPS positioning error caused by wheel slip and tree occlusion caused by road fluctuation in the park environment during the movement of unmanned vehicles.Relying on a single sensor and a simple data fusion method can not complete the positioning and navigation task in the complex park environment.Through data smoothing and improved adaptive weighted fusion algorithm,multisensor information fusion is realized to obtain more accurate odometer data.The global pose estimation obtained by fusing scan context and multi-sensor fusion algorithm improves the NDT algorithm,improves the accuracy and accuracy of point cloud matching,and realizes accurate positioning in the park environment.(3)In view of the difficulty of grid map generation caused by the ups and downs of the roads in the park,the ground points are separated by extracting the plane features of different road sections,and then the driveable areas are divided to ensure the normal driving of unmanned vehicles.Unmanned vehicles will encounter pedestrians and vehicles during operation.Relying on a separate lidar can not effectively and accurately judge the number and category of obstacles.Using the laser camera combined with target detection method to detect dynamic and static obstacles can effectively solve the problem of obstacle recognition.(4)Complete the performance test of unmanned vehicle positioning and navigation in the system.The environmental point cloud map is established by laser slam,and the effect of the improved adaptive weighted fusion algorithm and NDT algorithm in pose estimation is verified in the actual scene.The dynamic obstacle avoidance and point-to-point navigation performance of unmanned vehicle are tested.The final experimental results show that all indexes of the system meet the positioning and navigation requirements of unmanned cleaning vehicle.The thesis has 62 figures,17 tables and 83 references. |
PDF Full Text Request