| dc.description | Vehicle congestion is perhaps the biggest problem related to transportation that the inhabitants of big metropolis must face day by day, generating problems such as long commute periods, waste of productive time, air pollution, stress in drivers, overspending on fuel for vehicles, etc. Several solutions has been attempted to mitigate the problem, but so far this effort has not been enough. The work presented here pretends to attack this problem, going further with one of these solutions which consist in providing of additional intelligence to the traffic light controller systems in a street intersection, so they can administrate in a better way the traffic flow. This is achieved by taking advantage of the new embedded system technologies applied to IoT by monitoring and sharing in real time the traffic density in each one of the streets that converge in the intersection and based on this data, dynamically assign “time in green” proportionally to the. The proposed system is based on a multi-agent approach, in which each one of the traffic lights in an intersection is an individual agent which can communicate with the other agents through messages to exchange information( for example the traffic density). The proposed system uses a “leadership shared” scheme in which each one of the agents becomes the agent leader from time to time being this agent the one which calculates the lamp transition times in whole the system. This leader agent is selected in agreement with all the agents in the system, following either a pre-defined way or based on the traffic density reported by each agent. The proposed system was implemented using Intel Galileo boards (one per agent) creating a software stack which includes Yocto as OS, a Java Virtual machine, a multi-agent middleware called Jade and a multilayer software application. The system proposed is flexible enough to be adapted to any traffic light topology or any number of lamps in the traffic lights. This also supports the handling of over-traffic and emergency conditions, and it is able recover itself of any scenario of loss of communication or energy. The system was tested using a specific topology under different scenarios demonstrated that the behavior of the system works as expected. | |
