Wheel Slip Estimation Using a Neuromorphic Vision System (NVS) for Space Rovers

Abstract

Wheeled rovers navigating deformable and soft extraterrestrial terrains often face significant challenges related to wheel slip, which impacts navigation, traction control, and power consumption. Estimating terrain-induced slip in dynamic environments is particularly difficult due to unknown soil properties, low-light conditions, and the challenge of accurately measuring traversal speed.
This thesis presents a novel vision-based framework for real-time estimation of wheel slip, utilizing an RGB camera to capture terrain traces. These images are converted to grayscale for processing, where a filtering algorithm extracts relevant features and a Hough transform is employed to track terrain lines for slip ratio estimation. However, due to the limitations of RGB cameras in space applications—such as restricted sensor resources, limited power and memory, and poor performance under low-light conditions— they are not ideal for extraterrestrial environments. To address these challenges, a neuromorphic vision system (NVS) is adopted for slip estimation. The NVS offers high temporal resolution, which is critical for capturing rapid slip events typically encountered on planetary surfaces like the Moon.
Experimental validation using a single-wheel test rig demonstrated the framework’s ability to estimate slip ratios ranging from 2% to 90%, achieving over 94% accuracy in low-light environments and over 95% in well-lit conditions. Additionally, a field experiment was conducted using a real rover to validate the algorithm under operational conditions. The algorithm successfully estimated velocities with an accuracy of 95% in well-lit environments and 85% under low-light conditions. Furthermore, the framework was generalized to accommodate wheels without grousers. When tested with the neuromorphic vision sensor mounted on the front of the rover, the system achieved velocity estimation accuracies of 90% in well-lit settings and 84% in low-light environments.

Date of Award	2025
Original language	American English
Supervisor	Seneviratne Seneviratne (Supervisor)