Industrial-Scale Autonomous Vehicle Path Following by Feedback Linearized Iterative Learning Control
This work describes and demonstrates, through simulation and field trials, a technique for autonomous wheeled vehicle path following that uses iterative learning control (ILC) performed in a feedback linearized space to augment a base feedback linearization (FBL) path-following controller. The goal of ILC is to iteratively adjust steering rate inputs to account for unmodelled vehicle dynamics, environmental disturbances, and extreme path geometries. One fundamental advantage of this approach is that ILC can be used without having to employ approximate linearization at every time step, rendering the approach easily implementable and computationally inexpensive when compared with traditional approaches. The technique was validated by performing field trials using large industrial-scale autonomous underground mining vehicles. The presented work not only demonstrates the underlying technique in the field on commercial vehicles, but also proposes and validates a method for parallel speed learning, wherein the speed can be adjusted over subsequent learning trials to improve productivity. Finally, a method for pre-learning through simulation prior to deployment in the field is introduced in order to reduce initial path-following errors.