Autonomous vehicle obstacle detection and tracking

Almost all autonomous underwater vehicles require some manner of acoustic obstacle detection and avoidance mechanism to prevent vehicle damage due to collisions with uncharted obstacles. Modern forward looking multi-beam sonars use wide bandwidth pulses, fast sampling rates, and high bit count samples to form simultaneous beams having wide dynamic range and good target to clutter ratios. As a result, even moderate resolution sonars can generate sonar image data at rates of 10 Mbits per second or higher. This relatively large quantity of sonar data must be reduced to some sort of predicted physical environment ahead of the vehicle as part of the obstacle avoidance process. A data compression technique optimized for autonomous vehicles is described. A short video example of the algorithm working in real time on typical sonar data is presented