In this paper, we present a design of controller for vehicular automation based on mobile agent technology. In this approach, the control goal is conducted by coordinating active or default agents based on sensory feedback and vehicle conditions. A hosting mechanism determines which agent has the right to control the vehicle and make the decision. Those agents are divided into two groups: default agents which are developed by fuzzy rules, and active agents which are produced by neural networks. Fuzzy agents locate in the vehicle and neural network agents locate in remote server. Two types of agents have similar function but different performance, so those agents produced by neural network can transfer back to the vehicle after learning to replace default fuzzy agents according to mobile agent technology. The performance of the proposed technique is illustrated by simulation studies of a vehicle longitudinal control system