Design and Control of a New Omnidirectional Levelling System for Hilly Crawler Work Machines

To address the challenges posed by significant body inclination, suboptimal work quality, and safety concerns during agricultural machinery operations in hilly terrains, this study presents the design of an omnidirectional levelling system tailored for hilly crawler work machines. Initially, a structural configuration for an articulated omnidirectional levelling system employing a "three-layer frame" is proposed. Subsequently, the hydraulic circuit for levelling and its operational principles are elucidated. Furthermore, a sensitivity analysis is performed on the crucial structural parameters affecting omnidirectional levelling performance. To optimize these parameters, a multi-objective genetic algorithm (NSGA-II) is devised. Next, a novel composite control algorithm, denoted as QBP-PID (combining Q-learning, a BP neural network, and PID control), is proposed to enhance the adaptability of the PID controller. Real-time updates of PID control parameters are achieved through the BP neural network, with the incorporation of a Q-learning algorithm for online adjustment of neural network connection weights. The simulation results indicate that under QBP-PID control, the levelling time is reduced to 2.8 s for 20 degrees lateral levelling and 3.2 s for 25 degrees longitudinal levelling, with no observable overshoot compared to that of the PID and BP-PID controls. Finally, an omnidirectional levelling performance test is conducted, where the static test results align with the simulation findings. In dynamic test, the maximum body inclination angle of the crawler work machine is notably reduced, enabling swift levelling. The body inclination angle can be maintained within +/- 1.5 degrees on flat ground and after levelling, meeting the performance criteria for hilly areas and affirming the efficacy of the omnidirectional levelling system.