A case study of Machine Learning for Bio-Inspired Navigation

Contemporary navigation techniques, with which we’re all acquainted, are incredibly accurate but highly dependent on global navigation satellite systems, abbreviated GNSS. Although, GNSS is globally recognised, daily used, and thoroughly integrated in the quotidian life, for example integration in smartphones and smartwatches, it sometimes might not be available or have a weak signal.

Alternative navigation such as inertial navigation systems, also known as INS, are a reliable short-term solution. Based on a known starting point, which can be thought of as the point where GNSS is not available anymore, the INS is calculating position via gyroscopes and accelerometers among others. Because of accelerometers and gyroscopes drift, which is error accumulation from noise or other sources, the INS accuracy degrades rapidly over time raising new challenges.

My work is focused on developing artificial intelligence systems, which leverage techniques used by animals to navigate. It is a cross-disciplinary research between physics, biology, computer science, mathematics and electrical engineering. The aim is to mitigate the error accumulation, if not completely remove it, with a bit of luck.

My background as an artificial intelligence engineer working for Montel Group is what recommended me mostly for this project, as both positions include a keen eye on uncertainty. As of now, polarized light and how it is used for navigation by animals is the foundation of my project. The biggest success to date is creating a computer simulation of sun polarized light in the atmosphere and the evaluation of polarized light for navigational purposes. I can say that since I started this chapter of my life, I am enjoying life more because I understand it better, so let there be light!

If such an artificial intelligence system could be completed and used, the difference it would make would be of intense magnitudes. For example, considering current world events, navigating warzones to save hostages would be much safer and easier. Such a system would enhance the accuracy and reliability of alternative navigation systems, increase the autonomy of vehicles, enhance operation security and much more, which will put my partner, Raytheon in a competitive advantage in the market.

Back to: Centre for Doctoral Training in Distributed Algorithms