［Purposes］Because the existing planning models choose road links from the candidate road link set to build the dedicated lanes for autonomous vehicles, they cannot guarantee the connectivity of the established dedicated lanes for the autonomous vehicle. With the unconnected dedicated lanes, autonomous vehicles have to frequently enter and exit dedicated lanes increasing the probability of accidents and the travel time of autonomous vehicles. To address these issues, it presents a planning model to build the connected dedicated lanes for autonomous vehicles. ［Methods］The model is a two-level programming model, in which the upper problem is to minimize the system total travel time, and the lower problem describes the equilibrium assignment problem of the human-driven and autonomous vehicles under the traffic network with the connected autonomous vehicle dedicated lanes. Existing algorithms cannot solve the model since the candidate set to build the dedicated lanes is a path set. Thus, it developed an algorithm based on the simulated annealing algorithm to solve this model. ［Findings］By the numerical tests, one can find that the connected dedicated lane planning model can guarantee the connectivity of the dedicated lanes for autonomous vehicles. When the market penetration rate of the autonomous vehicle is low, the connected dedicated lane scheme can still effectively reduce the system total travel time. ［Conclusions］When autonomous vehicles have a high market penetration rate, human-driven vehicles should be only allowed to drive on the connected dedicated lanes for human-driven vehicles. This scheme can guarantee the right of way for human-driven vehicles, and the autonomous vehicles can also effectively take advantage of their potential to improve road performance and reduce the system total travel time.