Date of Award

2025

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering

Committee Chair

Avimanyu Sahoo

Committee Member

Farbod Fahimi

Committee Member

Dinh Nguyen

Research Advisor

Avimanyu Sahoo

Subject(s)

Robots--Motion, Robots--Control systems, Mobile robots, Reinforcement learning

Abstract

Autonomous wheeled mobile robots (WMRs) are widely used for safe operation in safety-critical systems, such as robotic visual inspection of confined spaces in energy infrastructure, warehouse automation, delivery robots, and autonomous vehicles. The operating environments for these safety-critical systems are often uncertain. Therefore, in such environments, it is essential for WMRs to reliably follow predetermined paths while effectively maintaining lane position and avoiding collisions. However, frequent observations required for control execution to account for environmental uncertainty result in increased sensing, computation, and energy costs. This necessity drives the research for safe and resource-aware trajectory-tracking control methods. Although several state-of-the-art trajectory tracking control schemes exist, these methods do not simultaneously address the challenges of safety and optimality under intermittently available sensing and computation. This research addresses the problem of optimal, safe trajectory tracking control for WMRs by developing a safe reinforcement learning (RL)-based trajectory tracking control framework integrated with event-based sensing and computation. The first part of the research reviews the state-of-the-art approaches for safe, resource-aware, and optimal control frameworks for WMRs in uncertain environments. It primarily focuses on 1) traditional control approaches for WMRs, 2) defining the rationale for selecting the control barrier function (CBF) as the safety certificate in the trajectory tracking control algorithm, and 3) the event-triggered control (ETC) that can reduce sensing and computation costs. In the second step, a near-optimal event-based sampling and optimal tracking control scheme under input constraints for WMRs is developed by extending an existing event-based RL-based control. The optimal trajectory tracking controller employs event-based adaptive dynamic programming and reinforcement learning to approximate the near-optimal controller under limited sensing and computation under input constraint. Numerical simulation results indicate a 61.2\% reduction in computation and sensing. In the third and final step, the event-based optimal trajectory tracking control is extended to incorporate safety by reformulating the cost function using CBF. The effectiveness of the proposed safe, event-triggered reinforcement learning-based control algorithm is validated through MATLAB-based numerical simulations in a lane-keeping scenario with safety constraints.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.