The emergence of autonomous vehicles can provide drivers with a chance to freely implement activities in the vehicles. On the basis of a representation of the in-vehicle activity utility of commuters in the autonomous vehicles, a multi-modal activity-based bottleneck model is proposed. In the proposed model, congestion interaction between regular and autonomous vehicles at the bottleneck is considered. The properties of the proposed model are analyzed. A continuous dynamic congestion tolling scheme is proposed for fully eliminating the bottleneck queue. The optimal step tolling scheme is designed to approximate the continuous dynamic tolling scheme, and its efficiency is evaluated in terms of the eliminated bottleneck queue. The results show that under the linear time-varying marginal activity utility, ignoring the in-vehicle activity utilities of commuters in the autonomous vehicles could lead to a biased estimation of departure time window of commuters during the morning commute.The commuters using the regular and autonomous vehicles sequentially depart from home in the morning, and there is the optimal market share of the autonomous vehicles that generates the lowest total social cost. Compared to the no-toll case, the queuing delay elimination rate of the optimal step tolls exceeds 〖KF(〗3〖KF)〗/3.