Publication Date

Spring 2024

Degree Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Engineering

Advisor

Wencen Wu; Mahima Agumbe Suresh ;Jun Liu

Abstract

In the exploration of mobile sensing agents such as robots equipped with sensors, these agents have the potential to measure unknown scalar fields, ranging from chemical concentrations to temperature fluctuations. One of the primary interests is the robots’ ability to gravitate towards the peaks or troughs of these fields, a phenomenon termed as “source seeking”. This behavior is invaluable in practical scenarios, like detecting chemical leaks or locating survivors after catastrophes. However, the challenge becomes significantly more pronounced when operating in turbulent flow fields, where the erratic nature of odorant plumes makes predictions and simulations highly unreliable. Interestingly, fruit flies exhibit a unique ability to navigate chaotic flows, which serves as the foundation of our research. Drawing inspiration from this natural adeptness of fruit flies, this research intends to devise algorithms for robots, initially focusing on singular robot source seeking before delving into collaborative multi-robot explorations. The primary objective is to develop search policy using Reinforcement Learning (RL) that guarantees convergence based on sensory feedback, all while accommodating the robots’ kinematic and dynamic constraints. Our methodology is anchored on evaluating data from real-world fruit fly experiments, using the dense trajectories near light sources to train models that replicate their behavior. Using a RL-based approach, robots act within a POMDP framework, guided by limited environmental insights. The efficacy of these controllers will be compared with current models to highlight the benefits of emulating fruit fly behaviors in robotic source seeking.

Download

Share

COinS