Behavior-Based Formation Control of H-SURF Robot Swarm through Relative States

Abstract

Autonomous Underwater Vehicles (AUVs) represent a critical component of exploring the largely uncharted 95% of the ocean. Inspired by the collective behaviours of fish schools, bird flocks, and swarms of bees, AUVs hold immense potential for deep sea exploration, discovery of marine species, and search and rescue missions. This project focuses on leveraging the Heterogeneous Swarm of Underwater Robotic Fish (H-SURF) to investigate and test control methods essential for these applications. This research delves into behaviour-based formation control using the relative states of HSURF robots. Implemented behaviours include fencing and milling using acoustic communication. Fencing behavior ensures agents stay within predefined boundaries, while milling makes the agents move around the boundary in clockwise or counter-clockwise manner. There are three boundaries created for testing the behaviours; circle, square, and Isotoxal star. Simulation results showed success with small deviations beyond the boundaries, where the error increases with the increase in number of agents. In order to evaluate the behaviours for experimenting, three methods of tracking were used and they are: 1) tracking system above water, 2) acoustic modem data, and 3) video analysis. Results showed that the tracking system did not have enough coverage of the pool surface, so it was used to evaluate the other methods, where the acoustic modem data proved to be accurate enough for assessing the behaviours unlike the video analysis. Successful experiments were carried out at the KU Marine Lab pool to assess the behaviors of 1, 2, and 3 agents individually, demonstrating effective results. The experimental results, despite the expected degradation due to data losses and system uncertainties, confirmed the simulated finding, both in qualitative behaviours and within quantitative expectations.

Date of Award	20 Apr 2024
Original language	American English
Supervisor	Federico Renda (Supervisor)