The trial of the TrafficGo project by the Huawei Research Center (Munich - DE) has come to an end.
Started in May 2019, it had the aim of testing on Turin traffic a traffic optimization technology, part of an Intelligent Trasnportation System ( ITS) solution called Traffic-Go, based on Artificial Intelligence algorithms.
Huawei is experimenting and testing the same technology in other European cities and obviously also in several Chinese metropolises, while Turin was chosen for the declination of the algorithm at European level and to be a reference point in the use of adaptive traffic optimisation systems.
More precisely, the local experimentation aimed at validating the solution proposed by Huawei on the traffic models of Turin under the supervision of 5T, the in-house company of the Municipality of Turin that manages the Mobility Centre of Turin. The validation was carried out in a simulated environment with no direct impact on traffic management, but no less complicated and important. Huawei believes in the need for a responsible use of artificial intelligence, exploiting simulated environments to validate the performance and robustness of its technologies in the laboratory before applying them in the real world. This is fully in line with the European Union's objectives for the development of artificial intelligence that can be trusted (also known as Trustworthy AI).
The experimentation was conducted in 3 phases:
- definition of the simulation scenario which consisted of 7 days of traffic, starting from a reconstruction of the road network and based on traffic measurements and traffic light plans provided by 5T;
- translation of all the procedures developed to transform the measurements of the traffic sensors installed in the streets of Turin, the traffic light plans and the files of the road network into two simulation scenarios.
- Execution of experiments to evaluate the performance of the traffic optimisation algorithms on the modelled scenarios compared with the real traffic light plans used, and interaction with 5T experts on the results obtained expressed in terms of average waiting time at the traffic lights, and quantity of CO2 produced by the simulated traffic.
The main value acquired by Huawei is the experience in traffic modelling and the complexities of generating a simulation scenario that is as consistent and realistic as possible. In addition, Huawei has confirmed that the algorithm can work well to optimise traffic in European scenarios, achieving significant performance in both heavy rush-hour traffic and lighter traffic conditions, providing better traffic optimisation than the compared baselines. The algorithm, very efficiently, senses the traffic and modifies the traffic light plan at each cycle. It does so by continuously learning the relationship between decisions made and results, adapting to perceived traffic conditions while respecting regulatory constraints. The solution is potentially compatible with existing ITS systems as it is based on the use of measurements from the same sensors, and allows the full exploitation of the computing power of the cloud, and the optimisation capability of modern artificial intelligence methods. The combination of these technologies opens up new applications and potential optimisations that will revolutionise the way traffic is managed in the years to come.
See the project sheet here.