Smart traffic framework based on dynamic mobile clusters

With the global trend towards urbanization, traffic control becomes an especially important problem. Existing `intelligent´ traffic control systems are usually of large scale, requiring high computation resources cost, especially equipped roads and traffic lights, and necessary legislations from governments. In this paper, we propose a framework resolves these issues by utilizing the processing power and sensing capabilities of smart devices, as well as relying on constrained optimal global traffic routing, with vehicle speeds, for managing the traffic. The framework constructs on-demand clusters of mobile smart devices, allowing for executing MPI, short-lived, parallel tasks. The tasks are coupled with the on-device sensors thereby decreasing sensed data to tasks communication overheads. The proposed framework requires low-cost coordinating servers, residing on a central cloud, and scales with the mobile devices. Preliminary results are obtained and analyzed based on a number of mobile devices forming a small cluster of heterogeneous mobile nodes. The results confirm the potential scalability of the mobile clusters for a typical optimal traffic control algorithm, and the utility of using standard cluster modeling techniques to predict their performance, making them amenable to standard optimizations.