Path Planning Algorithm Of Mobile Robot And Terrain Classification And Prediction Method

The development of mobile robots has shown huge practical application value and rich diversity.Among them,the UAV has the characteristics of lightness and agility,the wheeled crawler has excellent load capacity,and the leg-footed robots only need discrete footpoints in the actual movement process,so they have powerful obstacle crossing capabilities for driving in complex terrain.The environment information needed to support the actual operation of the mobile robot is the environment sensing system it carries,which usually coordinate the processing of the environment information by multiple sensors.By processing the environmental information obtained by the vision system,it provides a basis for solving problems such as mobile robot motion strategy,path planning,and obstacle avoidance algorithm.With the continuous development of mobile robots and their vision field,the research focus has transitioned from the realization of the structural design,motion stability analysis,and environmental information acquisition of mobile robots to how to improve the adaptability of mobile robots in complex environments,the ability of mobile robots to perceive environmental information and their autonomous initiative.In order to construct and optimize the vision system for mobile robots,this paper conducts a comprehensive and detailed study on the construction of environment perception system,path planning,terrain prediction,and terrain classification of mobile robots under complex environmental conditions.The main contents are as follows:(1)The ADFA*algorithm is proposed.This algorithm is suitable for grid maps and can be widely used for various mobile robots to solve their path planning problems.In order to improve the computational efficiency of the path planning algorithm,reduce the number of grids to be traversed,implement the obstacle avoidance strategy and limit the computational time cost,this paper improves on the traditional A*algorithm,and proposes the DA*algorithm that uses path splicing,the FA*algorithm that realizes obstacle avoidance by adding grid barrier levels,and the ADFA*algorithm that has more practical application value and time-consuming limitation are respectively proposed.The DA*algorithm uses the traditional A*algorithm to obtain the initial path at the initial moment and screens out the path nodes with compound standards on the initial path to segment the initial path.When the terrain information changes,the DA*algorithm does not need to re-plan the path from the initial starting point,but takes the path segmentation node closest to the target point as the starting point for path planning.Thereby greatly reducing the number of grids traversed by the A*algorithm to improve planning efficiency.The FA*algorithm is based on the A*algorithm and introduces the obstacle level F corresponding to each obstacle node to realize the expansion of the grid node where the obstacle is located.This method provides convenience for artificially regulating the obstacle avoidance safety level in special environments.Its advantage lies in the realization of obstacle expansion without changing the original path as much as possible to ensure the safety of the mobile robot body.ADFA*limits the heuristic part of the evaluation function by introducing parameters ? to achieve the purpose of controlling the time consumption of the algorithm.Simulation experiments in environments such as webots confirm that the number of grids traversed by ADFA*algorithm is significantly less than that of A*algorithm,and its search efficiency is more than 7 times that of A*algorithm(2)A map construction method based on the C-terrain terrain belt is proposed,and a terrain prediction method is proposed based on this.Based on the grid map for regional division,this paper proposes a fast and accurate terrain classification method C-Terrain by comparing the similar relationships between different regions.Based on the complete path planning,this method first obtains a set of orderly passing nodes at the initial moment.Then according to the choice of topographic zone and the parameter adjustment of the evaluation function,the ordered sequence of the influence function value is obtained.Finally,the regression method of machine learning is used to complete the prediction of the path and location terrain,and can realize the prediction of the unknown path and terrain.The experiment uses 10 sets of 50,000 frames of terrain data for prediction and simulation.The results prove that the accuracy of the C-Terrain terrain prediction method is above 90%,and the algorithm has high practical value.(3)The HMC terrain frame classification method is proposed.In the actual processing,a series of continuous terrain frames obtained by the airborne vision system will not undergo sudden changes.Based on this feature,this paper proposes a HMC(HMARF-MAP-CNN)terrain frame prediction classification method for the point cloud terrain frame classification problem.This method uses historical terrain data to classify and predict unknown terrain frames so that the mobile robot can perform gait switching and motion compensation.First,according to the storage structure characteristics of the raster map,the hidden Markov random field(HMRF)and the maximum posteriori(Maximum a Posteriori)are used to divide the terrain frame into flat terrain and rugged terrain.Then use CNN to filter the terrain frames belonging to the rugged terrain,which can be subdivided into rugged terrain and systematic-errors-caused.The terrain classification method can quickly and effectively realize the prediction and classification of terrain frames,has a sufficiently high classification accuracy,and can provide sufficient basis for the gait switching of the mobile robot in time.Simulation experiments show that the classification accuracy of HMC and HM methods both reach more than 89%without considering the training cost,and when processing complex terrain frame data,the classification accuracy of HMC both reaches more than 91%.It can be seen that the HMC composite terrain frame classification method proposed in this paper can effectively solve the classification problem of point cloud terrain data and has stronger stability.