This post is also available in: עברית (Hebrew)
Deep reinforcement learning (an aspect of machine learning) was used to generate new development enabling the landing of a fixed wing UAV in urban and maritime environments.
The very first UAV to perform a perched landing using machine learning algorithms has been developed in partnership with the University of Bristol, UK, and BMT Defence Services (BMT), a subsidiary of BMT Group Ltd.
The innovative development of a fixed wing aircraft that can land in a small or confined space has the potential to significantly impact intelligence-gathering and the delivery of aid in a humanitarian disaster.
According to BMT’s website, the 18-month research project was delivered as part of the UK Defence Science and Technology Laboratory’s (Dstl) Autonomous Systems Underpinning Research (ASUR) programme. BMT and Bristol University have demonstrated how the combination of a morphing wing UAV and machine learning can be used to generate a trajectory to perform a perched landing on the ground. The UAV has been tested at altitude to validate the approach and the team are working towards a system that can perform a repeatable ground landing.
Current UAVs are somewhat restrictive in that they have fixed and rigid wings, which reduces the flexibility in how they can fly. The primary goal of the work was to look at extending the operation of current fixed wing UAVs by introducing morphing wing structures inspired by those found in birds. To control these complex wing structures, BMT utilised machine learning algorithms to learn a flight controller using inspiration from nature.
Simon Luck, Head of Information Services and Information Assurance at BMT Defence Services comments: “Innovation is at the heart of everything we do at BMT and R&D projects provide us with the opportunity to work with our partners to develop cutting-edge capabilities that have the potential to revolutionise the way we gather information.”