In this paper we build an agent based evacuation model and use it to test a novel traffic control strategy called throttling. The evacuee agents travel from a source to a destination taking the dynamic shortest time path (total travel time depends on the distance to destination and the congestion level). Throttling involves closing a road segment temporarily when its congestion level reaches an upper threshold and opening it when congestion level falls below a lower threshold. Experimentation was performed by comparing the total evacuation time obtained with throttling to a base case (non-throttling) using a small test network and the more realistic Sioux Falls network. We found that throttling improves the total evacuation time significantly. To further test the effectiveness of our control strategy we compared it to contraflow on the test network and found the results to be comparable.