Limits of probabilistic safety guarantees when considering human uncertainty
From Murray Wiki
Jump to navigationJump to search
|Limits of probabilistic safety guarantees when considering human uncertainty
|Richard Cheng, Richard M. Murray and Joel W. Burdick
|Submitted, 2021 Conference on Decision and Control (CDC)
|When autonomous robots interact with humans, such as during autonomous driving, explicit safety guarantees are crucial in order to avoid potentially life-threatening accidents. Many data-driven methods have explored learning probabilistic bounds over human agents' trajectories (i.e. confidence tubes that contain trajectories with probability ), which can then be used to guarantee safety with probability . However, almost all existing works consider . The purpose of this paper is to argue that (1) in safety-critical applications, it is necessary to provide safety guarantees with , and (2) current learning-based methods are ill-equipped to compute accurate confidence bounds at such low . Using human driving data (from the highD dataset), as well as synthetically generated data, we show that current uncertainty models use inaccurate distributional assumptions to describe human behavior and/or require infeasible amounts of data to accurately learn confidence bounds for . These two issues result in unreliable confidence bounds, which can have dangerous implications if deployed on safety-critical systems.