Perching failure detection and recovery with onboard sensing

Abstract

Perching on a vertical surface carries the risk of severe damage to the vehicle if the maneuver fails, especially if failure goes undetected. We present a detection method using an onboard 3-axis accelerometer to discriminate between perching success and failure. An analytical model was developed to calculate acceleration differences for success and failure and set decision times. Two distinct decision times were shown to be effective, corresponding to properly engaging the gripper and overloading the gripper's capabilities. According to a machine learning feature selection algorithm, the maximum Z axis acceleration of the quadrotor and the presence of near-zero readings are the most relevant features within these two time frames. Using these features, the detection algorithm discriminated between success and failure with a 91% accuracy at 40 ms, and 94% at 80 ms. Real-time detection and failure recovery experiments with a 20 g quadrotor verify the detection method. An improved approach that combines various decision times correctly identified success/failure for all 20 trials with an average total falling distance of 0.8m during recovery. We discuss the feasibility of extending our method to other quadrotor platforms.