For the last 200,000 years, on a planet covered 70% by the sea, with only 5% of the oceans being explored by our species, underwater robots can be the best chance we have. Such robots should be fully autonomous with safety guarantees and robust behavior in cluttered, or even hostile, environments.
An ideal solution to the problem above is the Aqua2 AUV, which is a 6-legged agile amphibious robot with a unique propulsion system. But the robot's unique capablities of motion and autonomy come with some tradeoffs: Although a fair amount of studies in the past attempted to provide a complete hydro-dynamic analysis, due to its unique complexity, no such model has been produced. Moreover the limited computing power that is mostly occupied for Computer Vision and State Estimation processes, was prohibiting for more than a decade basic Motion Planning and thus true autonomy.
We introduce the first robust algorithm for safe navigation of such AUVs that utilizes the full mobility of the robot in clutterred 3-D environments using a geometric planning approach, without the need of a dynamics model or the excessive computation needs of other, state-of-the-art, approaches that are either assuming vertical relief or even they are bounded in 2-D.
The method operates both offline within a known map, or online within unknown environment that is observed online using stereo-cameras. Although a very limited bottom-facing field of view, the given challenge with underwater state estimation, and the limited computing power of the robot, a robust light-weight framework, based on prior work on underwater navigation and state estimation, is introduced.
The main goal for this project is to enable the Aqua2 navigate safely in previously unreachable for robots cluttered environments for inspection, mapping and monitoring. Shipwrecks is a class of environments we are particular interested since they combine all the challenges with underwater navigation, and at the same time, they are environments highly cluttered and unstructured, thus robust behavior of AquaNav in shipwrecks will prove scalability for simpler, more common, scenarios. There are 3 main directions which AquaNav is being developed:
Robust near-optimal behavior for a known map.
Robust localy optimal behavior for an unknown map utilizing the stereo cameras.
Robust navigation by producing motions that are maximizing real-time the quality of state-estimation. (Under development)
Robust monitoring and mapping by producing motions that are maximizing the visibiity of certain areas of interest. (Under development)