An important aspect of risk assessment for UAV flights is energy consumption, as running out of battery during a flight brings almost guaranteed vehicle damage and a high risk of property damage or human injuries. Predicting the amount of energy a flight will consume is challenging as many factors affect the overall consumption. In this work, we propose a deep energy model that uses Temporal Convolutional Networks (TCNs) to capture the time varying features while incorporating static contextual information. Our energy model is trained on a real world dataset and doesn’t require segregating flights into regimes. We showcase an improvement in power predictions by 35.6% on test flights when compared to a state-of-the-art analytical method. Once we have an accurate energy model, we can use it to predict the energy usage for a given trajectory and evaluate the risk of running out of battery during flight. We propose using Conditional Value-at-Risk (CVaR) as a metric for quantifying this risk. We show that CVaR captures the risk associated with worst-case energy consumption on a nominal path by transforming the output distribution of Monte Carlo forward simulations into a risk space and computing the CVaR on the risk-space distribution. Our state-of-the-art energy model and risk evaluation method helps guarantee safe flights and evaluate the coverage area from a proposed takeoff location.
Overview of the risk assessment pipeline