Decentralised Multi-intersection Congestion Control for Connected Autonomous Vehicles

This paper presents a decentralised mechanism for traffic control of connected autonomous vehicles in settings where multiple road intersections have to be managed and optimised. We propose a solution based on the distributed constraint optimisation approach (DCOP). We build upon state of the art algorithm for single-intersection management in order to manage congestion both across and within intersections. Furthermore, to solve the DCOP, we propose an improved node ordering policy for the Max-sum AD VP algorithm. Empirical evaluation of our model and algorithm demonstrate that our approach outperforms existing benchmarks by up to 32% in terms of average delay for both single and multiple intersection setup.