Date of Award

2024

Document Type

Open Access Master's Thesis

Degree Name

Master of Science in Electrical and Computer Engineering (MS)

Administrative Home Department

Department of Electrical and Computer Engineering

Advisor 1

Tan Chen

Committee Member 1

Jung Yun Bae

Committee Member 2

Hongyu An

Abstract

With NASA's ongoing efforts to establish a presence on the lunar surface to eventually move on to exploring mars, the development of intelligent robotic systems is more important than ever. The ability of robotics to explore hazardous and extreme terrain, coupled with long communication times from earth places an ever increasing need on more robust and efficient autonomy. Tethered robotics offer unique advantages to explore scientific targets both on the lunar and martian surfaces, capable of using their tether as a physical or metaphorical lifeline to allow the exploration of slopes, extreme dark regions, or areas in which wireless communication is ineffective. One such tethered robot, JPL's Axel rover, uses its tether both for power and physical stability on steep slopes. As it traverses, its tether comes into contact with the terrain, creating a difficult non-Markovian planning problem. This thesis explores further ways in which the success rate and efficiency of the planner can be improved, focusing on finding initial solutions. Two separate methods are used to improve the planner. The first biases the algorithm's sampling process to focus a portion of its approximation on sloped regions of the map, where the planner is expected to have more trouble finding a solution satisfying its stability constraints. The second re-configures the underlying Random Geometric Graph to form edges using a $k$-nearest neighbor approach, allowing longer edges to be evaluated. Both of these methods are thoroughly tested on a series of five maps, three of which are generated using a pipeline involving Blender and simulated sensors in Gazebo. These maps are used to evaluate four possible planner configurations in a comparison to the original work, added noise to perturb contact normals and rover pose, explore how the configurations scale to larger maps, and provide an example for a simple path over real-world terrain. Overall, the kNN Random Geometric Graph configuration is found to greatly improve the performance of the planner in most situations and raises the efficiency. It does identify more invalid solutions, indicating that a revision to the anchor prediction algorithms and static stability analysis may be needed. The slope biasing is found to marginally improve the planner in some cases, with the largest benefits occurring when the map is perturbed with noise. Finally, kNN is also found to scale much more effectively to the map size, offering incredibly large improvements over the R-disc configuration on a large map, and effectively matching its performance on a small map.

Creative Commons License

Creative Commons Attribution-Share Alike 4.0 License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 License.

Available for download on Monday, March 03, 2025

Share

COinS