Erwin Walraven

Traffic flow optimization

Traffic congestion causes important problems such as delays, increased fuel consumption and additional pollution. In this paper we propose a new method to optimize traffic flow, based on reinforcement learning. We show that a traffic flow optimization problem can be formulated as a Markov Decision Process. We use Q-learning to learn policies dictating the maximum driving speed that is allowed on a highway, such that traffic congestion is reduced. An important difference between our work and existing approaches is that we take traffic predictions into account. A series of simulation experiments shows that the resulting policies significantly reduce traffic congestion under high traffic demand, and that inclusion of traffic predictions improves the quality of the resulting policies. Additionally, the policies are sufficiently robust to deal with inaccurate speed and density measurements. HighlightsWe model a traffic flow optimization problem as a reinforcement learning problem.We show how speed limit policies can be obtained using Q-learning.Neural networks improve the performance of our policy learning algorithm.Resulting policies are able to significantly reduce traffic congestion.Our method takes traffic predictions into account and controls proactively.

Column Generation Algorithms for Constrained POMDPs

In several real-world domains it is required to plan ahead while there are finite resources available for executing the plan. The limited availability of resources imposes constraints on the plans that can be executed, which need to be taken into account while computing a plan. A Constrained Partially Observable Markov Decision Process (Constrained POMDP) can be used to model resource-constrained planning problems which include uncertainty and partial observability. Constrained POMDPs provide a framework for computing policies which maximize expected reward, while respecting constraints on a secondary objective such as cost or resource consumption. Column generation for linear programming can be used to obtain Constrained POMDP solutions. This method incrementally adds columns to a linear program, in which each column corresponds to a POMDP policy obtained by solving an unconstrained subproblem. Column generation requires solving a potentially large number of POMDPs, as well as exact evaluation of the resulting policies, which is computationally difficult. We propose a method to solve subproblems in a two-stage fashion using approximation algorithms. First, we use a tailored point-based POMDP algorithm to obtain an approximate subproblem solution. Next, we convert this approximate solution into a policy graph, which we can evaluate efficiently. The resulting algorithm is a new approximate method for Constrained POMDPs in single-agent settings, but also in settings in which multiple independent agents share a global constraint. Experiments based on several domains show that our method outperforms the current state of the art.