Real-time optimal control of quadrocopters using deep representations of the optimal state feedback

A major challenge in the field of control is to achieve reliable, aggressive, high-speed control of autonomous vehicles. In space, this may involve spacecraft that need to land under harsh conditions, or even – in an extreme scenario – negotiating asteroid debris fields at high speeds. On Earth, the exemplar task that draws most attention currently is high-speed autonomous flight of drones. The application of optimal control on board limited platforms has been severely hindered by the large computational requirements of current state-of-the-art implementations.In this work, we introduced and applied a deep neural network to directly map the robot states to the optimal control actions to overcome this limitation. The approach has been illustrated with high-speed flight of a drone with heavily constrained onboard processing, and can be applied to other platforms such as spacecraft that have similar restrictions.