betaFeedback

WorksEngineering

Information-Region MPC: Non-Iterative Distributed Control of Urban Traffic Signals

Abstract

Centralized model predictive control (MPC) exploits network-wide traffic information to obtain the optimal solution of the network-level control problem, but its computational burden grows rapidly with network size. Iterative distributed MPC reduces the size of local optimization problems, but requires repeated information exchange and coordination within each control update. To address these limitations, this paper proposes information-region MPC (IR-MPC), which approximates the first control input of centralized MPC using limited real-time data without iterative coordination. Based on historical traffic data, we quantify how perturbations at other intersections affect the centralized first control input of each intersection and use these couplings to construct a compact information region for every intersection. At each control update, each intersection then solves a local IR-MPC problem once using current data only from its information region. SUMO simulations on a real 28-intersection network show that IR-MPC with one processor per intersection computes its control inputs 9 times faster than centralized MPC and 15 times faster than a representative iterative decomposition-based distributed MPC method. IR-MPC reproduces 85% of the centralized control inputs exactly at the implemented one-second resolution, keeps 99.3% within 2s, and increases mean delay by only 0.03%.

Paper details

vlnc:W_2n345twm
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
arXiv nonexclusive license
Usage

views · 4

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Engineering

Claude

View statement
Looking for related papers…