This paper studies the variation of commute behaviors caused by unmanned vehicles from the perspective of activities. Considering the in-vehicle utility and the utilization efficiency of in-vehicle time, it formulates the activity-based utility function of commuters riding unmanned vehicles. Then it derives the departure and parking patterns, optimal dynamic bottleneck charge, and the model properties with constant and linear marginal-activity utility, respectively.These are compared with the corresponding results in the trip-based model.The results show that, with the constant marginal-activity utility,the early departure rate is an increasing function of in-vehicle activity utility and the utilization efficiency of the in-vehicle time, but the late departure rate reverses it. Meanwhile, with the linear marginal-activity utility, the number of early arriving commuters is not affected by the in-vehicle time loss. Still, it is positively related to the working start time. After implementing the optimal bottleneck toll, the trip-based model overestimates the upper bound of total toll revenue and underestimates the lower bound of total toll revenue. To maximize the net system utility, the government needs to adjust the parking density based on the parking social cost and the specific situation.