Chance-Constrained Iterative Linear-Quadratic Stochastic Games
This work addresses safety constraint satisfaction in multi-robot planning under uncertainty, particularly for autonomous driving, but it is incremental as it builds on existing iterative linear-quadratic methods with a new constraint-handling approach.
The paper tackles the problem of solving chance-constrained stochastic games for multi-robot planning under uncertainty, proposing the CCILQGames algorithm that uses augmented Lagrangian methods to avoid hand-tuned penalty weights, and it demonstrates safe and interactive strategies in autonomous driving scenarios like merge, intersection, and roundabout.
Dynamic game arises as a powerful paradigm for multi-robot planning, for which safety constraint satisfaction is crucial. Constrained stochastic games are of particular interest, as real-world robots need to operate and satisfy constraints under uncertainty. Existing methods for solving stochastic games handle chance constraints using exponential penalties with hand-tuned weights. However, finding a suitable penalty weight is nontrivial and requires trial and error. In this paper, we propose the chance-constrained iterative linear-quadratic stochastic games (CCILQGames) algorithm. CCILQGames solves chance-constrained stochastic games using the augmented Lagrangian method. We evaluate our algorithm in three autonomous driving scenarios, including merge, intersection, and roundabout. Experimental results and Monte Carlo tests show that CCILQGames can generate safe and interactive strategies in stochastic environments.