Simulations show that an eco-speed control algorithm can reduce fuel consumption up to 18 percent in the vicinity of signalized intersections.
A machine learning technique, Q-learning, fuels optimal decision-making.
Made Public Date


United States

A Conceptual Demonstration of Self-Learning Eco-Speed Control at Signalised Intersection

Summary Information

Traffic idling at intersections wastes 2.8 billion gallons of gasoline annually in the United States, in addition to contributing to air pollution. Signalized intersections are identified as a main reason behind idling fuel consumption due to stop and go conditions related to signal changes. In order to address this problem, this study presents a novel eco-speed control algorithm to assist with fuel-efficient driving at signalized intersections.


The proposed algorithm is designed to derive the fuel-optimal driving trajectory for an individual vehicle by eliminating idling in the area of an isolated signalized intersection. The algorithm uses three main input data sources: the status of the traffic signal, vehicle speed, and vehicle position from an onboard GPS receiver. Using these data, the algorithm then determines the eco-friendly speed trajectory while obeying traffic constraints. The algorithm uses Q-learning, a self-learning, intelligent agent to optimize driving speed to minimize fuel consumption. Q-learning is a machine learning technique used for searching for an optimal action-selection policy for the finite Markov decision process. Initially, the model has no knowledge and no expectation of outcomes. As trials continue, data are collected, and after each trial the model is better at estimating the desired outcome.


The eco-speed control algorithm demonstrated that the fuel consumption can be reduced up 18 percent and vehicle idling time can also be reduced.

A significant benefit of the self-learning algorithm is that the control system can directly learn from past experience under various environmental conditions.
Deployment Locations