A Distributed Gradient Approach for System Optimal Dynamic Traffic Assignment

TL;DR

This paper introduces a distributed gradient-based method for solving system optimal dynamic traffic assignment problems, enabling decentralized computation with high-quality solutions in finite iterations.

Contribution

It proposes a novel distributed algorithm that decomposes SODTA into local sub-problems with neighbor communication, improving scalability and solution quality.

Findings

Achieves solutions with at most 5% optimality gap

Converges within a finite number of iterations

Effective across different demand scenarios

Abstract

This study presents a distributed gradient-based approach to solve system optimal dynamic traffic assignment (SODTA) formulated based on the cell transmission model. The algorithm distributes SODTA into local sub-problems, who find optimal values for their decision variables within an intersection. Each sub-problem communicates with its immediate neighbors to reach a consensus on the values of common decision variables. A sub-problem receives proposed values for common decision variables from all adjacent sub-problems and incorporates them into its own offered values by weighted averaging and enforcing a gradient step to minimize its objective function. Then, the updated values are projected onto the feasible region of the sub-problems. The algorithm finds high quality solutions in all tested scenarios with a finite number of iterations. The algorithm is tested on a case study network under different demand levels and finds solutions with at most a 5% optimality gap.