AUV Multi-agent Formation Control

Sammanfattning

This thesis focuses on creating a multi-agent leader-follower formation control system for Autonomous Underwater Vehicles (AUVs), implementing an Extended Kalman filter (EKF) to provide accurate estimations of leader vehicles, enabling the vehicles to maintain their relative positions within the formation. With the increasing use of AUVs, research has focused on improving localization methods, as the system is limited by low bandwidth and high latency of acoustic channels, which are commonly used for data transmission. This filter was then tested in a simulated environment where two leader AUVss communicated with a follower AUV to intermittently share position data via acoustic pings. Multiple obstructive scenarios with obstacles blocking the acoustic pings were also tested to compare the effectiveness of the formation control and the accuracy of the filter. These tests also showed that in scenarios where the follower vehicle was affected by currents or experienced communication cut-offs, all errors remained bounded, and the error consistently remained around 1% of the traveled distance, which corresponded to approximately 1200 meters. In cases where the vehicles exhibited turning movements, the transient errors were shown to be reduced once the turns were completed. Overall, the results demonstrate that combining EKF with Proportional–Integral–Derivative (PID) is an effective method for accurately guiding AUVs in simulated environments and provides a backing on which real world testing can be justified.