Risk-Bounded Autonomous Vehicle Off-road Navigation Under Wheel Slip Uncertainty

Abstract

Wheel slip prediction is essential for safe navigation and optimizing the trajectory planning of ground vehicles including rovers, especially when traversing off-road on unpaved surfaces such as sand, gravel, or mud. However, calculating tire slippage precisely is cumbersome due to numerous sophisticated processes of measuring physical parameters related to the wheel-soil interaction. Most prior studies focused on developing slip prediction models suited for rovers or differential-drive robots, leaving car-like robots relatively overlooked. Moreover, prior works used complex handcrafted techniques that required complex terrain-specific parameters. This paper presents a data-driven deep-learning approach for predicting ground vehicle wheel slip in uneven and unpaved terrain. First, extensive data collection is carried out with an advanced simulator to construct a sufficiently descriptive dataset (504,000 samples) capturing various terrains, speed ranges, slopes, and maneuvers. Then, considering the close correlation between the terrain type and wheel slippage, we propose a lightweight convolutional neural network (CNN), coined as TerrainNet, for accurate terrain classification. The produced slip model to plan a free slip-induced path. Moreover, we proposed a modification of the Rapidly Exploring Random Trees (RRT) path planning algorithm coined as 4-DOF S-RRT. The proposed algorithm tries to avoid the slip-induced area by checking the predicted slip value and its distribution (i.e., slip uncertainty) from traversing over two nodes. The approach maintains the slip value under a certain bounded threshold and is feasible in a real-world application. The simulation results indicate that the proposed CNN can accurately discriminate the terrain (mean accuracy > 99%), enabling precise wheel slip estimations with the employed machine learning models (average root mean square error < 0.03). Moreover, the proposed path planning model successfully integrated the slip model and generated a feasible path while avoiding obstacles and slip-risk terrain given a certain threshold with 49 out of 50 success rates.

Date of Award	Dec 2022
Original language	American English
Supervisor	Majid Khonji (Supervisor)