Researchers from MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) have developed a process that allows users to easily design hybrid drones of different sizes and shapes
A hybrid drone design merges the fuel efficiency and speed of a fixed-wing aircraft with the vertical take-off and landing of a rotor aircraft. Hybrid drone control systems often have to be designed from scratch for every drone, which is difficult, expensive and time consuming.
The MIT CSAIL method uses neural networks to create more adaptable control systems. CSAIL developed a new component in the neural network controller that keeps track of the discrepancies between simulation and real-world scenarios that allow the controller to adapt its output command. The system does not store any modes and can switch from hovering to gliding by simply updating the drone’s target velocity.
“Our method allows non-experts to design a model, wait a few hours to compute its controller, and walk away with a customized, ready-to-fly drone,” said Jie Xu, an MIT CSAIL grad student. “The hope is that a platform like this could make more of these more versatile ‘hybrid drones’ much more accessible to everyone.”
Researchers integrated the systems into the CAD program OnShape to allow users to first select and match drone parts from a data set. The design of the drone was then inputted into a training process through a simulator that tests its flight performance.
The next steps are to further increase the drone’s maneuverability by improving its design and to work on the model to account for complex aerodynamic effects between the propeller’s airflow and the wings.