Comparing Context-Aware Routing and Local Intersection Management

In multi-agent routing, there is a set of mobile agents each with a start location and destination location on a shared infrastructure. An agent wants to reach its destination as quickly as possible, but conflicts with other agents must be avoided. We have previously developed a single-agent route planning algorithm that can find a shortest-time route that does not conflict with any previously made route plans. In this paper, we want to compare this route planning approach with non-planning approaches, in which intersection agents determine which agent may enter an intersection next, and where the agent will subsequently go (given its destination). When making these decisions, the intersection agents use only locally observed traffic information. Our experiments show that context aware routing produces more efficient results in case no incidents disrupt the execution. However, in the face of unexpected incidents, the performance of the intersection management policies proves very robust, while context-aware routing only produces good results when coupled with effective plan repair mechanisms.Software Computer TechnologyElectrical Engineering, Mathematics and Computer Scienc