Automatic Trigger Speed for Vehicle Activated Signs using Adaptive Neuro fuzzy system and ClassificationRegression Trees

Vehicle activated signs (VAS) are speed warning signs activated by radar when driver speed exceeds a pre-set threshold, i.e., the trigger speed. In order to be able to operate the sign more efficiently, it is proposed that the sign be appropriately triggered by taking into account the prevalent road and traffic conditions. This study presents the use of adaptive neuro-fuzzy inference systems (ANFIS) and classification and regression trees (CART) to predict the trigger speed of the VAS by using a historical speed data. The speed data is first explored and clustered by using a selforganizing map (SOM). Input vectors for simulation composed of time of day, traffic flow and standard deviation of mean vehicle speeds whereas the output vector consists only of vehicle speeds in the 85 percentile. The two models examined in this study were tested with historical speed data collected in Sweden during a period of one week and their performance was compared with Multi-layer perceptron (MLP). The results show that CART is reliable for predicting trigger speed for vehicle activated signs. However, compared to MLP and ANFIS, CART has superior performance than the other algorithms in terms of accuracy and complexity. Keywordsvehicle activated signs; adaptive neuro-fuzzy inference systems; classification and regression tree; selforganizing maps; trigger speed.