Game Theoretic Planning for Self-Driving Cars in Competitive Scenarios

Abstract

We propose a nonlinear receding horizon gametheoretic planner for autonomous cars in competitive scenarios with other cars. The online planner is specifically formulated for a two car autonomous racing game in which each car tries to advance along a given track as far as possible with respect to the other car. The algorithm extends previous work on gametheoretic planning for single integrator agents to be suitable for autonomous cars in the following ways: (i) by representing the trajectory as a piecewise-polynomial, (ii) incorporating bicycle kinematics into the trajectory, (iii) enforcing constraints on path curvature and acceleration. The game theoretic planner iteratively plans a trajectory for the ego vehicle, then the other vehicle until convergence. Crucially, the trajectory optimization includes a sensitivity term that allows the ego vehicle to reason about how much the other vehicle will yield to the ego vehicle to avoid collisions. The resulting trajectories for the ego vehicle exhibit rich game strategies such as blocking, faking, and opportunistic overtaking. The game-theoretic planner is shown to significantly out-perform a baseline planner using Model Predictive Control which does not take interaction into account. The performance is validated in high-fidelity numerical simulations, in experiments with two scale autonomous cars, and in experiments with a fullscale autonomous car racing against a simulated vehicle.

Pursuit-evasion Game for Nonholonomic Mobile Robots With Obstacle Avoidance using NMPC

“…In [15] several information patterns are reviewed and then Best response is used to compute the Nash equilibrium for two quadrotor drone racing. The work was improved in [16] where sensitivity enhanced iterated best response algorithms was used to solve for the approximate Nash equilibrium in the space of feasible trajectories and applied to a car-like vehicle. In [17], the concept of iterated best response and model predictive control was combined to solve for an agile interaction between two ground vehicles modeled in a semi-stochastic formulation.…”

Potential iLQR: A Potential-Minimizing Controller for Planning Multi-Agent Interactive Trajectories

“…To find Nash equilibria of the interaction game, in [13], a hierarchical decomposition of the underlying game into strategic and tactical games was proposed. Iterative best response algorithms were developed for capturing interactions in racing problems and autonomous driving settings [14,15,16]. In [17], iterative dynamic programming in Gaussian belief space was used to solve for equilibria of a game-theoretic continuous POMDP.…”

Lightweight Object Detection Ensemble Framework for Autonomous Vehicles in Challenging Weather Conditions

“…Leveraging computer vision for self-driving cars has evolved with the expanding requirements and research in the field and is now spread across several tasks, including vehicle detection, anomaly detection, trajectory prediction, object classification, path planning, collision avoidance, and modeling traffic rules [1,2]. As most of these systems are usually tested under simulations, the development and training under complex scenarios can be simulated using a variety of techniques, including modeling traffic using inspiration from the theory of multiagent systems, blocking and overtaking scenarios using RC cars, and an autoencoder trained with generative adversarial costs coupled with a recurrent neural network transition model [8][9][10][11].…”