An Iterative LQR Controller for Off-Road and On-Road Vehicles using a Neural Network Dynamics Model

TL;DR

This work addresses robust trajectory tracking for wheeled robots operating across both off-road and on-road environments. It combines neural-network learned dynamics with an Iterative Linear Quadratic Regulator (ILQR) within an MPC-like framework to compensate for model imperfections. The approach is validated on two platforms, the off-road Warthog and the on-road Polaris GEM e6, using multiple reference trajectories and showing feasible, constrained control. Together, these results demonstrate that learned dynamic models can be effectively integrated with ILQR to achieve accurate, robust tracking across diverse operating conditions.

Abstract

In this work we evaluate Iterative Linear Quadratic Regulator(ILQR) for trajectory tracking of two different kinds of wheeled mobile robots namely Warthog (Fig. 1), an off-road holonomic robot with skid-steering and Polaris GEM e6 [1], a non-holonomic six seater vehicle (Fig. 2). We use multilayer neural network to learn the discrete dynamic model of these robots which is used in ILQR controller to compute the control law. We use model predictive control (MPC) to deal with model imperfections and perform extensive experiments to evaluate the performance of the controller on human driven reference trajectories with vehicle speeds of 3m/s- 4m/s for warthog and 7m/s-10m/s for the Polaris GEM

Figures (9)

• Figure 1: Warthog
• Figure 2: Polaris GEM e6
• Figure 3: Figure shows bicycle mode of Polaris GEM e6 and error state w.r.t a reference trajectory
• Figure 4: Polaris GEM e6 Circular trajectory response
• Figure 5: Polaris GEM e6 Oval trajectory response