Navigation is a challenging problem, particularly in the underwater marine environment. The maximum visibility in pure water is 80 m but such clarity is rarely present in nature, limiting the utility of vision-based navigation techniques to much shorter distances in clear waters. A further challenge is the lack of visual features in open waters. Using an underwater polarization video system inspired by the mantis shrimp eye, we have previously shown that it is possible to bypass this limitation and perform celestial navigation without direct observation of the sun. The heading and elevation of the sun can be inferred from polarization angle measurements of the scattered, in-water light field. Our proof-of-concept system achieved an accuracy of 61 km in global positioning, or 0.38° in heading. To reduce the material, computational, and energy costs of the system, we have applied the biological design principles of sparsity and matched filtering. That is, we sense and compute the minimum information necessary for specific levels of accuracy and our sensors are tuned to the most informative set of signals in the environment. This work brings us closer to a practical realization of a new method for long-distance navigation in underwater vehicles, without the need to surface, and shows how biological design principles can ease requirements for real-time and resource-constrained systems.
|