On Infusing Reachability-Based Safety Assurance Within Probabilistic Planning Frameworks for Human-Robot Vehicle Interactions

Abstract

Action anticipation, intent prediction, and proactive behavior are all desirable characteristics for autonomous driving policies in interactive scenarios. Paramount, however, is ensuring safety on the road -a key challenge in doing so is accounting for uncertainty in human driver actions without unduly impacting planner performance. This paper introduces a minimally-interventional safety controller operating within an autonomous vehicle control stack with the role of ensuring collision-free interaction with an externally controlled (e.g., human-driven) counterpart. We leverage reachability analysis to construct a real-time (100Hz) controller that serves the dual role of (1) tracking an input trajectory from a higher-level planning algorithm using model predictive control, and (2) assuring safety through maintaining the availability of a collision-free escape maneuver as a persistent constraint regardless of whatever future actions the other car takes. A full-scale steerby-wire platform is used to conduct traffic weaving experiments wherein the two cars, initially side-by-side, must swap lanes in a limited amount of time and distance, emulating cars merging onto/off of a highway. We demonstrate that, with our control stack, the autonomous vehicle is able to avoid collision even when the other car defies the planner's expectations and takes dangerous actions, either carelessly or with the intent to collide, and otherwise deviates minimally from the planned trajectory to the extent required to maintain safety.

Roundabout Crossing With Interval Occupancy and Virtual Instances of Road Users

“…One of the most difficult issues in this scenario is handling vehicles that maneuver from the inner ring of the roundabout to the outer ring, which is the ring used by the AD vehicle in our method. In practice, it is very challenging to predict a lane change maneuver, especially when vehicles attempt to make a lane change with nudging [21]. In [25] we proposed three different strategies for handling the navigation of multiple vehicles inside a two-lane roundabout.…”

“…In a mixed traffic environment with AD and MD vehicles (in this paper an MD vehicle can be a car or any road user in the drivable space), avoidance sets can be computed with reachability theory [11], [21], [39]. The authors in [28], [29] implemented a collaborative motion planning algorithm that cooperates with MD vehicles.…”

Highly Parallelized Data-Driven MPC for Minimal Intervention Shared Control

“…The semi-autonomous vehicle paradigm is distinct from the concept of a fully autonomous self-driving cars as the autonomy's goal is not to take full control of the vehicle, but instead to act as a guardian or intelligent co-pilot [4], intervening on the person's control when deemed necessary to ensure safety. For this reason, there is a growing line of work that follows the minimum intervention principle in the semi-autonomous vehicle domain [25]. For example, Schwarting et al [28,29] describe a parallel autonomy framework that develops control trajectories for semi-autonomous vehicles that minimize deviation from userinput and achieve task-specific metrics like road following and contour tracking.…”

“…return u r 6: end procedure result" [3,25,28]. By developing a shared control algorithm that adheres to the MIP we maximize the influence of the human partner's control at each given moment.…”

“…The majority of related work requires hand-written (and potentially complex) models of the system and control dynamics [3,25,28]. In this work, we instead use a representation that is simple (i.e., linear) regardless of the underlying dynamics and can be learned from data when the model is not known a priori.…”