RESUMO
Unmanned aerial vehicles (UAV) rely on a variety of sensors to perceive and navigate their airborne environment with precision. The autopilot software interprets this sensory data, acting as the control mechanism for autonomous flights. As UAVs are exposed to physical environment, they are vulnerable to potential impairments in their sensory mechanism. Their real-time interactions with the actual atmosphere make them susceptible to cyber exploitations as well, where sensory data alterations through counterfeit wireless signals pose a significant threat. In this context, sensor failures can result into unsafe flight conditions, as the fault handling logic may fail to anticipate the context of the issue, allowing autopilot to execute operations without necessary adjustments. Untimely control of sensor failures can result in mid-air collisions or crashes. To address these challenges, we created Biomisa Arducopter Sensory Critique (BASiC) dataset, a state-of-the-art resource for UAV sensor failure analysis. The BASiC dataset comprises 70 autonomous flight data, spanning over 7 hours. It encompasses 3+ hours of (each) pre-failure and post-failure data, along with 1+ hour of no-failure data. We selected the ArduPilot platform as our demonstration aerial vehicle to conduct the experiments. By engineering Software in the Loop (SITL) parameters, we effectively executed sensor failure test simulations. Our dataset incorporates six representative sensors failures which are critical to UAV operations: global positioning system (GPS) for precise aerial positioning, remote control for communication with the ground control station (GCS), accelerometer for measuring linear acceleration, gyroscope for rotational acceleration measurement, compass providing heading information, and barometer for maintaining flight height based on atmospheric pressure data. The availability of the BASiC dataset will benefit the research community, empowering researchers to explore and experiment with state-of-the-art deep learning models by tailoring them for time series signal analysis. It may also contribute in enhancing the safety and reliability of mission-critical autonomous UAV flights.
RESUMO
Autonomous flights are the major industry contributors towards next-generation developments in pervasive and ubiquitous computing. Modern aerial vehicles are designed to receive actuator commands from the primary autopilot software as input to regulate their servos for adjusting control surfaces. Due to real-time interaction with the actual physical environment, there exists a high risk of control surface failures for engine, rudder, elevators, and ailerons etc. If not anticipated and then timely controlled, failures occurring during the flight can have severe and cataclysmic consequences, which may result in mid-air collision or ultimate crash. Humongous amount of sensory data being generated throughout mission-critical flights, makes it an ideal candidate for applying advanced data-driven machine learning techniques to identify intelligent insights related to failures for instant recovery from emergencies. In this paper, we present a novel framework based on machine learning techniques for failure prediction, detection, and classification for autonomous aerial vehicles. The proposed framework utilizes long short-term memory recurrent neural network architecture to analyze time series data and has been applied at the AirLab Failure and Anomaly flight dataset, which is a comprehensive publicly available dataset of various fault types in fixed-wing autonomous aerial vehicles' control surfaces. The proposed framework is able to predict failure with an average accuracy of 93% and the average time-to-predict a failure is 19 s before the actual occurrence of the failure, which is 10 s better than current state-of-the-art. Failure detection accuracy is 100% and average detection time is 0.74 s after happening of failure, which is 1.28 s better than current state-of-the-art. Failure classification accuracy of proposed framework is 100%. The performance analysis shows the strength of the proposed methodology to be used as a real-time failure prediction and a pseudo-real-time failure detection along with a failure classification framework for eventual deployment with actual mission-critical autonomous flights.