A study on automatic control of wheel loaders in rock/soil loading

Abstract

Wheel loaders are widely used in mines, construction projects, and waste processing fields due to their merits of high mobility, remarkable operational flexibility, and relative low capital cost. Automatic control of the loading process will achieve high productivity, solve the problem of shortage of experienced operators, release human workers from hazardous working environments, and reduce maintenance cost associated with machine abuse. Existing research on wheel loader control only automate the manipulator mechanism while leave the locomotion system to human operators. The research in this dissertation focuses on the development of a digging controller that can simultaneously control bucket motion and vehicle advancement, and can adapt its parameters in different digging conditions. In this controller, a digging task is decomposed into three sequentially implemented operations: attacking, crowding, and scooping; each of them can be carried out with different strategies. An ideal trajectory is pre-generated based on the physical property of the wheel loader and rock pile, and is generally followed during the digging process. Human operators' practice in regulating vehicle speed and coordinating bucket motion and vehicle advancement is intensively investigated, refined into control rules, and implemented with intelligent and conventional control methods. Machine features are analyzed in depth and incorporated into control algorithm design. The investigation on the non-trivial time-delay characteristic of the manipulator hydraulic system leads to the building of dynamic models and the application of predictive control principle in digging decision-making. The feature of small working space of wheel loader was considered in bucket trajectory generation and instant bucket motion planning. The research on wheel slip occurrence leads to a detection algorithm being developed and a slip restrain strategy being presented. Self-adaptation mechanism is constructed in which trajectory tracking-quality is applied for evaluating digging control performance, statistical analysis is applied for improving the accuracy on bucket penetration depth prediction, and machine-learning technique is applied for revising the membership function of fuzzy consequent variables. Simulation experiments show that the controller works effectively in rock piles with different conditions, from fine sand and soil to big fragmented rock pieces.