Online Controller Setpoint Optimization for Traction Control Systems Applied to Construction Machinery

Abstract

The development of a suitable traction control system for off-road heavy machinery is complicated by several different factors, which differentiate these machines from typical on-road systems. One such difficulty arises from the fact that they are often operated on ground conditions which can vary widely and rapidly. Due to this, traction control systems designed for these vehicles must be robust to a large array of surface types, and they must be capable of reacting quickly to significant changes in those types. In order to accomplish this, this paper proposes an online parameter optimization technique suitable for tuning the setpoint of a control system to maximize the tractive potential of a construction vehicle in real time. The traction control principle itself is based on selectively braking wheels which are slipping. It also attempts to account for the interactions of the transmission systems that deliver power from the engine to the wheels. This research uses a wheel loader as a reference machine for assessing controller performance. Drawing on previous work in simulation and controller design, a system model was developed which incorporates the vehicle dynamics of the machine as well as the behavior of the electrohydraulic brakes. This system model was leveraged to understand the effect of different optimization schemes on the performance of the traction control. The self-tuning algorithm is based on a compound optimization method utilizing both a system identification component and a parameter tuning component. The first part optimizes the model parameters to fit it as well as possible to measured slip-friction data. Based on the results of this, the second part draws from theories of wheel traction to maximize a balance of pushing force and traction effectiveness. The result is a method which can achieve the proper setpoint based on real-time data describing the ground condition. This system was run first in simulation and then on a modified vehicle system. In both cases, the algorithm allows the controller to find better setpoints to improve the traction control performance online.