GP-based Runtime Planning, Learning, and Recovery for Safe UAV Operations under Unforeseen Disturbances

Autonomous vehicles are typically developed and trained to work under certain system and environmental conditions defined at design time and can fail or perform poorly if unforeseen conditions such as disturbances or changes in model dynamics appear at runtime. In this work, we present a fast online planning, learning, and recovery approach for safe autonomous operations under unknown runtime disturbances. Our approach estimates the behavior of the system with an unknown model and provides safe plans at runtime under previously unseen disturbances by leveraging Gaussian Process regression theory in which a model is continuously trained and adapted using data collected during the autonomous operation. A recovery procedure is event-triggered any time a safety constraint is violated to guarantee safety and enable learning and replanning. The proposed framework is applied and validated both in simulation and experiment on an unmanned aerial vehicle (UAV) delivery case study in which the UAV is tasked to carry an a priori unknown payload to a goal location in a cluttered/constrained environment.

I. Introduction

Autonomous unmanned aerial vehicles (UAVs) and in particular multi-rotor vertical take-off and landing (VTOL) vehicles are rapidly finding their way into our society especially for transportation/delivery applications. In fact, thanks to their design and capabilities, they provide fast transportation between different locations and as demonstrated recently, they can be used to carry packages and even for medicine delivery. However, real-world operations bring several challenges, often not known a priori: state estimation uncertainties such as deprecated localization in GPS denied environments, external disturbances like wind, changes in model dynamics and system aging, component failures, and variation in payload distribution are some example of such factors that could lead to undesired conditions like the loss of stability and even collisions if not properly considered at design time or at runtime.