Queue-Arrival Fusion for Safety-Constrained Adaptive Control of Isolated Traffic Signals

Abstract

Adaptive traffic signal control at isolated intersections remains challenging under fluctuating demand, particularly when control strategies rely exclusively on either queue length or arrival rate as the primary decision variable. Queue-based approaches effectively respond to accumulated backlog but lack anticipation of emerging inflow, whereas arrival-rate-based strategies react to short-term demand intensity while potentially overlooking persistent congestion. This paper proposes a lightweight queue-arrival fusion based adaptive signal control framework that explicitly integrates both indicators within a unified green-time adaptation mechanism while strictly enforcing safety and operational constraints. The controller operates using only locally measurable traffic information and does not require prediction models, learning mechanisms, or inter-intersection coordination. Microscopic simulations conducted in SUMO demonstrate that the proposed fusion strategy outperforms both queue-only and arrival-rate-only adaptive controllers. Compared with queue-based control, average waiting time is reduced by 28.0 percent, average queue length by 36.4 percent, throughput increases by 5.9 percent, and multi-cycle clearance failure decreases by 58.3 percent. Relative to arrival-rate-based control, delay is reduced by 23.4 percent and clearance failures by 50.0 percent. These results demonstrate that explicitly fusing accumulated backlog and near-term arrival intensity yields consistent efficiency and stability improvements over single-indicator adaptive strategies.