Adaptive Traction for Optimal Mobility for Heavy Duty Vehicles

Abstract

This thesis addresses the critical need for developing prediction-based control allocation strategies for autonomous or manually operated vehicles within low-speed application area such as construction and mining sites. A prediction horizon focused in this thesis is approximately 50 meters. In the pursuit of safe and energy-efficient control, it is essential to harness the potential of multiple traction actuators, which traditionally operate re-actively. This project seeks to optimize these systems using predictive algorithms, given that drivers often lack the knowledge required to operate them effectively. Furthermore, the timely responsiveness of actuators is of critical importance in demanding situations. The current practice involves manual control of traction actuators, such as differential locks and electronically controlled air suspension, based on drivers’ real-time observations. However, this approach is often sub-optimal, as it does not fully utilize the capabilities of these systems. To address this issue, the thesis centers on automating these traction actuators, leveraging predictive road data. It assumes the availability of upcoming road data, including road profile and predicted friction data for the next 50 meters. The primary objective is to develop an optimal control strategy that maximizes traction while ensuring adequate steering margin. To achieve this, the thesis initially delves into understanding how these actuators influence traction and steering. Subsequently, a rule-based control allocation model is developed in MATLAB and Simulink, which is then tested with a comprehensive vehicle simulation model across various test cases. The research also extends to practical implementation. The control allocation logic is transferred to real-world conditions using real-time systems, specifically the MicroAutoBox II, on a physical truck. Impressively, the developed control function provides results in almost real-time, with a response time of only approximately 1000 milliseconds. While this computational time may be considered too high for safety-critical functions in some contexts, it remains adequate for the specific function under scrutiny, which is focused on predicting the upcoming 50- meter road conditions. In conclusion, the thesis presents a comprehensive approach to enhance traction using differential locks and axle load distribution strategy. By automating traction actuators based on predictive road data and optimizing control strategies, this research contributes to realizing safer, more energy-efficient autonomous driving systems.