Offroad autonomous vehicle development and model-based adaptive robust control strategy

Autonomous driving has seen increasing applications in off-road environments, driven by advancements in sensing and control technologies. To operate reliably under severe off-road conditions, a robust and adaptable off-road platform is essential. Off-road environments, in particular, present unique challenges, including low traction, uneven terrain, rapidly changing slopes, and high vibration levels. Model-based controllers have the potential to improve control performance in challenging off-road scenarios significantly. Preliminary work focused on simulator development and control architecture design for an electric vehicle prototype. Motivated by platform limitations and off-road driving requirements, the research was extended to develop an off-road vehicle platform, the R-Gator. A redesigned modular system architecture was implemented on the R-Gator, accompanied by a high-fidelity simulator that enables flexible and safe testing. Building on this upgraded platform, three key research efforts are presented to address the challenges of environmental disturbances in off-road path-tracking control. First, vehicle dynamics were identified using Dynamic Mode Decomposition with Control (DMDc) to enable model-based control, with noise reduced by a Savitzky–Golay filter. The identified model was validated against a Least Squares Estimation (LSE) baseline and actual vehicle responses under multi-input conditions. A Linear Quadratic Regulator (LQR) was then designed using the identified model for vehicle control. Second, to handle slope variations in off-road environments, a real-time system identification method combining Set Membership Estimation (SME) and Least Squares Estimation (LSE) was developed. The updated model was incorporated into a Slope-aware Adaptive Model Predictive Controller (SAMPC) to reduce slope-induced path-tracking disturbances. The framework was validated in simulation and real-world tests, improving performance over a standard MPC. Third, an Adaptive Tube Model Predictive Controller (ATMPC) was developed for complex off-road environments with unmodeled disturbances. Terrain classification enabled online model selection, while slope, slip ratio, and vibration indices were incorporated for adaptive dynamics and stability constraints. The controller was tested in both simulated multi-terrain settings and real-world experiments, showing improved tracking and stability over a standard MPC.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Mechanical Engineering

Grantor

University of Illinois Urbana-Champaign

Year dc:date

2025

Author dc:creator

• Zhang, Jiaming

Contributors dc:contributor

• Sreenivas, Ramavarapu S.
• Dullerud, Gier
• Krishnan, Girish
• Hsiao-Wecksler, Elizabeth T.

• Off-road Autonomous driving
• Simulator development
• System identification
• Adaptive Robust Model Predictive Control

Statement dc:rights

• Copyright 2025 Jiaming Zhang

Language dc:language

en

Handle dc:identifier

https://hdl.handle.net/2142/132673

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/132673