Design And Research Of Integrated Perception And Planning System For Car-like Robot

With the aging population,rising labor costs and other issues becoming increasingly prominent,the demand for intelligent mobile robots equipped with different sensing devices in the fields of logistics and security has surged.Such robots mainly move across regions on unknown indoor and outdoor roads,but the current navigation system is not well-suited to cross-region road application scenarios,including cross-regional global environment map construction and maintenance difficulties;motion planning is extremely dependent on the global path,and it is particularly difficult to apply to car-like robots with poor flexibility,and the system architecture is complicated and the maintenance cost is high.Therefore,an integrated perception and planning system for car-like robot scheme suitable for road scenes is proposed,including lightweight environment perception model design,adaptable motion planning model design,and integrated system experimental research,to achieve autonomous and stable movement of car-like robots in road scenes.The specific research work of the thesis is as follows:(1)Design and research of car-like robot environment perception model.Comprehensively analyze the characteristics of roads and obstacles in motion scenes,design a laser-vision combined perception scheme,use feature recognition,segmentation and fitting methods,establish obstacles,indoor and outdoor road models,and theoretical models for robot pose calculation;Considering factors such as compactness and simplicity,the composite configuration is selected as the robot equivalent geometric model,and the model parameters are optimized.Finally,according to the obstacle-roadrobot pose relationship,a complete lightweight motion planning map model is obtained.(2)Design and research of car-like robot motion planning and optimization model.By comprehensively analyzing the inherent relationship between the planning map model and the motion control instruction sequence,a motion planning model is established.A hybrid potential field model is designed,which can adaptively generate a safe initial path based on a planning map.The path constraint is converted into an unconstrained penalty function,and the least square method and the quasi-Newton method are used to jointly solve the problem,and the optimal path is obtained through optimization based on the initial path.The speed constraint is converted into a penalty function,the optimal speed analytical solution is obtained,and the optimal speed profile coupled with the optimal path is generated.Based on the Ackerman geometric model,the optimal speed profile is converted into a motion control sequence.The theory and examples are used to analyze the influence of the parameters on the planning results,and the correctness of the model principle is verified.(3)Car-like robot perception and planning system integration and experimental research.According to the requirements of the scene,a car-like robot platform are designed and built,and the key parameters calibration are completed through the "error amplification" experiment.Collect indoor and outdoor scene information through RPLIDAR lidar and ZED camera,design its software system according to the perception model,and the feasibility and reliability of obstacles and indoor and outdoor road modeling methods are verified.According to the motion planning model,the dynamic simulation experiments of straight and arc scenes are designed in MATLAB,and the motion planning model has good comprehensive performance is verified.Finally,the perception and planning system is integrated on the platform and tested in indoor and outdoor straight road scenarios.The experimental phenomena and data verify the reliability and practicability of the integrated perception and planning system studied in this paper.