Research On Travel Trajectory Planning And Tracking Control Of Loaders

Articulated wheel loader driverless and intelligent technology is the key development direction clearly pointed out in the Construction Machinery Industry "14th Five-Year Plan",which is an important means to achieve efficient and green construction of future loaders.At the same time,with the increasing cost of manpower and training difficulties,construction machinery enterprises are increasing the demand for intelligent equipment,especially in high temperature,high dust,high altitude and other difficult and high working conditions and monotonous reciprocal operating environment,construction machinery enterprises increasingly hope that through the unmanned intelligent construction machinery to complete the required operation.However,at present however,the current level of construction machinery equipment intelligence is still low,limited by intelligent planning and control methods,still remote control construction machinery as the main direction of development.Compared with the driverless system of passenger vehicles.Firstly,articulated wheel loaders usually operate under unstructured road conditions such as dirt roads,muddy roads,gravel roads,etc.,without lane line guidance,and cannot copy the passenger vehicle road detection algorithm for environment perception.Secondly,the articulated wheel loader usually works in a relatively closed area with reciprocal loading and short-distance transport operations,which cannot replicate the path decision and planning algorithm of passenger vehicles.Finally,since the loader is an articulated structure,its front and rear vehicles are controlled by the articulated structure for steering,which cannot replicate the tracking control algorithm of passenger vehicles.Therefore,if the unmanned driving of the articulated wheel loader is realized,the problems of environment recognition,trajectory planning and trajectory tracking applicable to the loader need to be solved,and for the above problems,the following research contents are mainly carried out in this paper:1.analyze and study the principles of current mainstream semantic segmentation methods,and select a Deep Lab V3+ semantic segmentation method applicable to the working environment of the loader.Under the assumption of flat ground plane and fixed camera height,the distance of surrounding obstacles is solved based on the ground feature point map.The obstacle area is expanded by the obstacle expansion method to complete the map environment construction.2.Under the condition that the environment map is known,the trajectory planning is carried out by using the improved ant colony algorithm.The concept of obstacle inflection point is proposed,and the search range is expanded by establishing the starting pointobstacle inflection point-end point relationship model.For the heuristic information,the idea of target distance is introduced to design the heuristic information calculation method;for the state transfer rule,the inflection point grid is used instead of the neighborhood grid for transfer probability control;for the path optimization,the 3 times B spline curve is used for the path smoothing process to realize the trajectory planning.The simulation tests show that the improved algorithm shortens the path length by 4.31%,1.192% and 52.70%,the search steps by 82.36%,84.38% and 91.55%,and the operation time by-189.84%,38.79%and 38.29% compared with the traditional algorithm,and there is no excess turn.3.The articulated wheel loader is the object,and the dynamics state space model is established based on the transverse dynamics model of the loader,considering the body coordinates and state deviation.By introducing the reversing mechanism in the control prediction model,the model is switched to achieve the purpose of continuous control of forward and reverse driving.4.With the pre-sighting path as the reference,the longitudinal vehicle speed and articulation angle as the control input,and the lateral position error and heading angle error as the control output,the kinetic prediction model is established.With vehicle speed,articulation angle and acceleration as constraints,the objective function is transformed into a quadratic programming problem to establish a loader model predictive trajectory control system to realize trajectory tracking control.The comparison experiments with different pre-targeting distances show that the model predictive trajectory control system based on the pre-targeting path can control the lateral deviation within 0.05 m when setting the pretargeting distance P=0.6m,and the average speed of steering motor is reduced by 6.42%,which is significantly better than the traditional model predictive trajectory control system for this test bench;The effectiveness of the system in "V" working condition and "T" shoveling working condition is also verified,which shows that the vehicle can be continuously tracked under this working condition,and the feasibility and effectiveness of the method described in this paper are verified.