Proceedings of the 2014 Industrial and Systems Engineering Research Conference Y. Guan and H. Liao, eds. Hierarchical En-Route Planning under the Extended Belief-Desire- Intention (E-BDI) Framework Sojung Kim, Young-Jun Son, Ye Tian, Yi-Chang Chiu The University of Arizona Tucson, AZ, USA C. Y. David Yang Federal Highway Administration McLean, VA, USA Abstract En-route planning is a dynamic planning process to find the optimal route (e.g., shortest route) while driving. The goal of this paper is to mimic a realistic drivers’ en-route planning behavior under the situations with incomplete information about road conditions using the Extended Belief-Desire-Intention (E-BDI) framework. The proposed E- BDI based en-route planning is able to find a new route to the destination based on the predicted road conditions inferred by drivers’ own psychological reasoning. A main challenge of such a detailed E-BDI model is a high computational demand needed to execute a large scale road network, which is typical in a big city. To mitigate such a high computational demand, a hierarchical route planning approach is also proposed in this work. The proposed approach has been implemented in Java-based E-BDI modules and DynusT® traffic simulation software, where a real traffic data of Phoenix, Arizona is used. To validate the proposed hierarchical approach, the performance of the en-route planning modules under the different aggregation levels is compared in terms of their computational efficiency and modeling accuracy. The validation results reveal that the proposed en-route planning approach efficiently generates a realistic route plan with individual driver’s prediction of the road conditions. Keywords Agent-based simulation, En-route planning, Belief-desire-intention, Hierarchical route planning. 1. Introduction Route planning by searching the optimal route on a road network given pairs of the origin and destination, is one of the major problems in the transportation and supply chain management applications [1]. To achieve accurate prediction and analysis of the traffic system, drivers’ route planning behavior s have been extensively studied considering individual driver’s own preferences (e.g., road safety, traffic volume, willingness to pay) [2]. Recently, Kim et al. [3] has proposed a route planning approach under the Extended Belief-Desire-Intention (E-BDI) framework [4], which has allowed us to mimic the realistic drivers’ en-route planning behavior with various preferences. The en-route planning approach (theme of this paper) based on such an E-BDI framework can generate a dynamic route plan under a more realistic driving environment (e.g., drivers’ learning and interactions while driving) [3]. One of the major difficulties of the E-BDI based en-route planning, however, is high computational requirements, especially for a large scale road network, which is typical in a big city. In order to infer the conditions of each road (or link), the E-BDI based en-route planning needs to execute its underlying algorithms multiple times, such as Bayesian Belief Network (BBN) which is a probabilistic model to handle uncertain perception and reasoning processes of a human under dynamic environments (e.g., perception of travel time on a road network) [5]. One way to resolve such a computation issue is to adopt a hierarchical route planning approach, which is known as efficient for speeding up the route planning [1]. The hierarchical route planning is an aggregation approach by dividing an