Formation control of autonomous underwater vehicles presents a key challenge due to the limitations of underwater acoustic communication. Traditional approaches require frequent information exchange about the vehicle status, which is not feasible in the underwater environment due to low bandwidth, long time delays, and low data transmission reliability. Underwater acoustic communication achieves only a few bytes per second, making it impossible to continuously track the status of the leader vehicle in the formation.
We developed a formation control system that is based on prediction of leader states with an extended Kalman filter (EKF). This system enables formation maintenance during long message exchange intervals. It uses a leader-follower approach, where follower vehicles maintain a prescribed offset from the leader.
We designed an EKF based on Fossen's dynamic model of underwater vehicles with a state vector that includes the position, velocity, accelerations, and heading of the leader vehicle. We formulated the state transition function that accounts for the hydrodynamic drag and inertia of the BlueROV2 vehicle. The key contribution is the integration of a proportional controller into the EKF prediction step, which enables the prediction of turns based on direction change commands.
We implemented the control system in ROS Noetic using the DAVE simulation environment. We developed a ROS package for simulating acoustic communication, which includes realistic delays based on Mackenzie's equation and acoustic modem specifications. We conducted testing on a triangular polygon with different formations (loose, line, V-formation) at various intervals of communicating the leader vehicle status from 20 to 240 s and under various environmental conditions.
Results show that the EKF successfully predicts the trajectory of the leader vehicle for all tested intervals. The formation control system maintains acceptable formation accuracy for communication intervals up to 90 seconds with distance errors below 10 m and angle errors below 0,5 rad. The errors increase at longer intervals due to the accumulation of the EKF prediction error, but the basic formation geometry remains preserved. Quality analysis of formation control under different environmental conditions revealed a nonlinear relation with the communication interval of reporting the status of the leader vehicle.
The limitations of the formation control system include a unidirectional message exchange approach, which creates a critical point of failure, and the possibility of formation breakup in the event of partially successful command receiving. Despite these limitations, the developed control system represents an important contribution to underwater robotics, as it enables formation control at lower communication frequencies than traditional approaches. This leads to lower energy consumption and reduced load on communication channels. The research enables the practical use of the algorithm for oceanographic research, hydrographic measurements, and underwater defense. Additionally, it provides a foundation for further development of autonomous underwater vehicles with improved energy efficiency and reduced acoustic communication activity.
|
