S. Spiller, Y. Zweiri
Orion Technology Group, LLC,
United States
Keywords: Robotics, Attitude determination, Dropout method, Deep learning, Multirotor UAV
Summary:
Orion CONTROL DNNA is an innovative sensor fusion technique being integrated into the Orion AIM artificial intelligence mission system that enables unmanned aerial systems to dynamically auto-tune flight controller parameters during flight based on current environmental conditions, improving system stability. Precise state estimation is needed for unmanned aerial vehicles to fly with a high degree of stability. Accurate state estimation is considered challenging due to the inaccuracies associated with the measurements of the onboard commercial-off-the-shelf (COTS) Inertial Measurement Unit (IMU) and the immense vibration of the vehicle’s rotors that makes these measurements suffer from issues like; large drifts, biases and unpredictable noise sequences. These issues cannot be significantly tackled using classical estimators and an accurate sensor fusion technique needs to be developed. A deep neural network (DNN) is trained using a deep-learning-based technique to identify the associated measurement noise models and filter them out. Dropout technique is adopted for training the DNN to avoid overfitting and reduce the complexity of nets computations for onboard processing, enabling flight controller dynamic tuning and improved flight characteristics. DNNA has been incorporated into Orion CONTROL software and is running systems as small as the 169g AI/ML Orion X4 nUAS.