Research On The Identification Of Bulk Stacked Material And The Evaluation Method Of Shoveling Difficulty

The road surface of engineering vehicles is complicated,the working environment is harsh,in the process of operation,the operator is very easy to fatigue,affecting the efficiency of work and there are safety hazards.It is of great practical significance to carry out research on unmanned engineering vehicle technology to reduce operator labor intensity,reduce operational accidents and improve operational efficiency.As a typical engineering vehicle,the loader involves many technologies in the process of realizing autonomous operation,among which the autonomous shoveling technology is the key to realize unmanned loader.Identification of bulk material and evaluation of the difficulty of shoveling are the basic prerequisites for the autonomous shoveling operation of the loader.According to the type of material and the degree of difficulty of shoveling,the corresponding shoveling strategy is formulated and a reasonable shoveling trajectory is planned to prevent the loader from various abnormal working conditions in the shoveling process.At the same time,material type identification and shoveling difficulty assessment are directly related to the intelligent adjustment of engine（or motor）output power,hydraulic system control and driving strategy.In view of the above problems,this paper takes the unmanned loader as the carrier to study the identification of bulk material accumulation and the evaluation method of shovel difficulty.This research is based on the National Natural Science Foundation of China project "Key Technology Research on Autonomous Shoveling of Unmanned Loaders"（No.51875233）,and the specific research contents are summarized as follows.（1）Based on an overview of the development process of unmanned loaders and existing research results of target detection technology at home and abroad,the research content and general technical route of this paper are proposed.（2）For the identification of bulk pile materials,firstly,real images of five pile materials under multiple scenes and non-uniform lighting conditions are collected,and the data set is expanded and processed by image processing techniques.From the viewpoint of detection accuracy and detection rate,YOLOv4 target detection network is selected to perform migration learning for the five kinds of stacked materials,removing the region proposal step and combining all calculations into one stage.applications.（3）For the assessment of the ease of shoveling of bulk materials,the process of autonomous shoveling by wheel loaders is first analyzed,and the main factors affecting the ease of shoveling are clarified,and the material shoveling stage is jointly determined based on multi-sensor information.The pressure changes of the movable arm cylinder in the insertion stage and the turning stage of the wheel loader are selected to indirectly reflect the changes of shoveling resistance encountered during autonomous shoveling,and the pressure data are processed by the segmented least squares fitting algorithm,and the four indicators of maximum pressure,pressure change rate,operation time length and time change rate of the movable arm cylinder in this stage are taken,and then the central values obtained by clustering are equally distributed on the four diagonal lines of the radar map by the clustering algorithm.The radar map method is used to comprehensively evaluate the shoveling difficulty of the stacked materials.（4）A real-time evaluation method of the ease of excavation of the identifiable pile material is proposed,which is a solution for real-time and efficient evaluation of the ease of excavation of the bulk pile material by fusing the neural network with the radar map information and matching the material type with the radar map information.It is a solution for real-time and efficient assessment of the ease of shoveling of bulk materials.In this paper,we build a material identification and excavation ease assessment system to evaluate the excavation ease of five types of bulk materials,train a material detector with neural networks,and verify the robustness of the system under complex environmental conditions to provide technical support for the autonomous excavation operation of wheel loaders.