Recently, with the trend of world-wide urbanization, some of the accompanying problems are getting serious, including road traffic congestion. To deal with this problem, city planners now resort to the application of the latest information and communications technologies. One example is the adaptive traffic signal control system (e.g. SCATS, SCOOT). To increase the throughput of each main intersection, it dynamically adjusts the traffic light phases according to real-time traffic conditions collected by widely deployed induction loops and sensors. Another typical application is the on-board vehicle navigation system. It can provide drivers with a personalized route according to their preferences (e.g. shortest/fastest/easiest), utilizing comprehensive geo-map data and floating car data. Dynamic traffic assignment is also one of the key proposed methodologies, as it not only benefits the individual driver, but can also provide a route assignment solution for all vehicles with guaranteed minimum average travel time. However, the non-recurrent road traffic congestion problem is still not addressed properly. Unlike the recurrent traffic congestion, which is predictable by capturing the daily traffic pattern, unexpected road traffic congestion caused by unexpected en-route events (e.g. road maintenance, an unplanned parade, car crashes, etc.), often propagates to larger areas in very short time. Consequently, the congestion level of areas around the event location will be significantly degraded. Unfortunately, the three aforementioned methods cannot reduce this unexpected congestion in real time. The contribution of this thesis firstly lies in emphasizing the importance of the dynamic time constraint for vehicle rerouting. Secondly, a framework for evaluating the performance of vehicle route planning algorithms is proposed along with a case study on the simulated scenario of Cologne city. Thirdly, based on the multi-agent architecture of SCATS, the next road rerouting (NRR) system is introduced. Each agent in NRR can use the locally available information to provide the most promising next road guidance in the face of the unexpected urban traffic congestion. In the last contribution of this thesis, further performance improvement of NRR is achieved by the provision of high-resolution, high update frequency traffic information using vehicular ad hoc networks. Moreover, NRR includes an adaptation mechanism to dynamically determine the algorithmic (i.e. factors in the heuristic routing cost function) and operational (i.e. group of agents which must be enabled) parameters. The simulation results show that in the realistic urban scenario, compared to the existing solutions, NRR can significantly reduce the average travel time and improve the travel time reliability. The results also indicate that for both rerouted and non-rerouted vehicles, NRR does not bring any obvious unfairness issue where some vehicles overwhelmingly sacrifice their own travel time to obtain global benefits for other vehicles.