Competitive Driving of Autonomous Vehicles

Abstract

This paper addresses the issue of autonomous competitive yet safe driving in the context of the Indy Autonomous Challenge (IAC) simulation race. The IAC is the first multi-vehicle autonomous head-tohead competition, reaching speeds of 300 km/h along an oval track modeled after the Indianapolis Motor Speedway (IMS). We present a racing controller that attempts to maximize progress along the track while avoiding collisions with opponent vehicles and obeying the race rules. To this end, the racing controller first computes a race line offline. During the race, it repeatedly computes a small set of dynamically feasible maneuver candidates, each tested for collision with the opponent vehicles. It then selects a collision-free maneuver that maximizes the progress along the track and obeys the race rules. Our controller was tested in a 6-vehicle simulation, managing to run competitively with no collision over 30 laps. In addition, it managed to drive within a close range of the leading vehicle during most of the IAC final simulation race.

A Physics-Driven Artificial Agent for Online Time-Optimal Vehicle Motion Planning and Control

“…The development of online minimum-time motion planning and control techniques for autonomous vehicles is a complex and partly unsolved problem [1]. Simulations have been setup [2], [3], [4], [5], [6] to validate the algorithms in controlled environments, before the (often risky and expensive) experimental testing [7], [8], [9], [10], [11], [12]. The planning and control tasks become even more involved when most of the vehicle parameters are not known.…”

“…The longitudinal load transfer was modeled in [23] to improve a G-G diagram. 3 The impact of model fidelity for minimum-time trajectory optimization was analyzed in [25]. Some researchers used machine learning tools to augment NMPC tracking controllers.…”

“…The term plant is used with its meaning in control theory (i.e., the combination of actuators and process) 2. The G-G-v diagram is a 3D graphical representation of the performance envelope of a vehicle, where the x, y, and z axes correspond to the maximum vehicle's longitudinal acceleration, lateral acceleration, and longitudinal velocity 3. The G-G diagram is a 2D graphical representation of the performance envelope of a vehicle, where the x and y axes correspond to the maximum vehicle's longitudinal and lateral accelerations.…”

Online Reachability Analysis and Space Convexification for Autonomous Racing

“…Over the last several years, autonomous racing has actively been pursued as a strategy to explore edgecase scenarios in autonomous driving [16]. Racing scenarios present unique challenges with respect to navigating high speeds and multi-agent interactions.…”

Pursuit Path Planning for Multiple Unmanned Ground Vehicles Based on Deep Reinforcement Learning

“…These methods plan paths based on local observations, making them preferable for large-scale vehicle systems. In the Independent Actor-Critic (IAC) [19] approach, each agent learns a decentralized policy and critic locally to generate a local experience at each timestep. Another architecture, called Centralized Training with Decentralized Execution (CTDE) [20], learns decentralized policies in a centralized manner, which allows vehicles to access each other's observations and unobservable extra-state information during training.…”