Abstract
In real-world settings, from woking in manufacturing plants to self driving on highway, robots empowered by Machine Learning (ML) models are tasked with complex, dynamic tasks that demand high levels of precision and adaptability. The reliability of these systems hinges on the perception capabilities of ML model, making uncertainty quantification methods vital. Conformal prediction is a user-friendly paradigm for creating statistically rigorous uncertainty sets/intervals for the predictions of such models. It ensures that robots can effectively assess and respond to varying conditions with safe and trustworthy actions, reducing the risk of errors and enhancing overall system performance.
The purpose of this research project is to experimentally validate the effectiveness conformal prediction in object detection of a control algorithm on a ground robot platform.
Document Type
Article
Class Name
Electrical and Systems Engineering Undergraduate Research
Date of Submission
12-8-2023
Recommended Citation
Li, Yifei, "Experimental Validation of Uncertainty Quantification Methods for Robot Perception" (2023). Electrical and Systems Engineering Undergraduate and Graduate Research. 21.
https://openscholarship.wustl.edu/eseundergraduate_research/21