Available at: https://digitalcommons.calpoly.edu/theses/3063
Date of Award
6-2025
Degree Name
MS in Electrical Engineering
Department/Program
Electrical Engineering
College
College of Engineering
Advisor
Xiao-Hua Yu
Advisor Department
Electrical Engineering
Advisor College
College of Engineering
Abstract
This thesis presents a biologically inspired multi-agent control system for real-time path planning and task allocation in a dynamic and obstacle-laden underwater environment, specifically for teams of Autonomous Underwater Vehicles (AUVs). Traditional methods including classical heuristic algorithms and AI-based approaches often fail to effectively adapt to dynamic environments or require trained policies for each specific task space. To address these issues, this work proposes an approach that integrates Glasius Bio-Inspired Neural Networks (GBNNs) and a Collaborative Discrete Artificial Bee Colony (CDABC) algorithm, along with a ”gradient-of-neighbors” path planning algorithm based on local neural activity gradients, to produce smoother and more efficient trajectories and task assignments that adapt in real time to changing environments.
Simulation results with ten different random environments demonstrate that the new multi-agent control system policies significantly reduce mission completion time — achieving average improvements of 65.1% as compared to conventional GBNN-based systems, with a trade-off of increased average travel distance of 19.5%. More tests with more realistic environments will be performed to fully evaluate the performance of the proposed approach.