Session: 07-11-03 Mobile Robots and Unmanned Ground Vehicles III

Paper Number: 71179

Start Time: Tuesday, 06:40 PM

71179 - Stochastic Predictive Control for Crash Avoidance in Autonomous Vehicles Based on Stochastic Reachable Set Threat Assessment 

Unexpected hazardous situations frequently arise on the roads, which lead to crashes on the road. According to a recent report from the Association of Road International Safe Travel, approximately 1 million people die worldwide in road accidents every year. Around 30 million people suffer from injuries resulting in long-term disabilities. This could be attributed to drinking and driving, getting sleepy or reckless driving on the roads.  In an NHTSA study, it was reported that around 94% of car crashes could be attributed to driver errors and misjudgments. Therefore, it is crucial to detect anomalous driving behaviors on the road and react quickly and appropriately in a near-crash situation. Detecting a threat in advance and generating a fallback trajectory for crash avoidance are some of the major challenges faced by autonomous vehicles. Stochastic Model Predictive Control (SMPC) approaches have recently proven to be highly effective for controlling systems in highly uncertain environments. This paper presents a fast, proactive decision-making approach for autonomous vehicles using SMPC to avoid crashes with other vehicles on the road. Two problems are addressed in this study: (1) Assessing threat to the autonomous vehicle from the surrounding vehicles and (2) Generating a safe trajectory for the autonomous vehicle in case of a future predicted threat. First, the problem of threat assessment is solved by modeling the surrounding vehicles using driver-behavior-based Stochastic Reachable (SR) sets. These SR sets allow formulating a surrounding driver's possible future movements depending on their predicted driving behavior. Three different driver behaviors are modeled: normal, aggressive, and drowsy. Each driver's behavior type poses a different threat to the ego vehicle. The second problem of generating a safe trajectory is solved using a proactive decision-making approach. A proactive decision-making approach exploits the surrounding human-driven vehicles' intent to assess the future threat, which helps generate a safe trajectory in advance, unlike reactive decision-making approaches that do not account for the surrounding vehicles' future intent. The crash avoidance problem is formulated as a SMPC problem where chance-constraints are specified to account for uncertainty in the surrounding vehicle's motion. These chance-constraints always ensure a minimum probabilistic safety of the autonomous vehicle by keeping the probability of crash below a predefined risk parameter. This paper proposes a tractable and deterministic reformulation of these chance-constraints using convex hull formulation for a fast real-time implementation. A safe area is computed at each time instant using the probabilistic reachable sets of the surrounding vehicle and the predefined risk parameter. This safe area is further formulated using convex hull representation, which allows it to be represented using linear inequality constraints. The controller's performance is studied for different risk parameters used in the chance-constraint formulation. The proposed approach will be tested on simulations for three different road scenarios: Oncoming vehicle, cut-in scenario, and intersection scenario. We expect our algorithm to ensure the safety of the autonomous vehicle in each hazardous situation in each road scenario.

Presenting Author: Vanshaj Khattar Virginia Tech

Authors:

Vanshaj Khattar Virginia Tech
Azim Eskandarian Virginia Tech