In order to guide CAVs' autonomous lane-changing decisions and yielding behaviors, and to explore the dynamic characteristics of their autonomous behaviors, a system dynamics model of vehicle-to-vehicle interactions was established. The model determines the spatial threshold for CAVs, controls the target cruising speed of CAVs in bottleneck areas, adjusts the traffic density of road vehicles in bottleneck areas, and improves the efficiency of heterogeneous traffic flow. MATLAB numerical simulations were conducted to validate the dynamic model for CAVs' autonomous decision-making behaviors. The results show that, compared to traditional traffic optimization models, incorporating autonomous behaviors factors for CAVs under lane-changing and yielding conditions reduces the average delay of vehicles in bottleneck areas and the queue length by approximately 20%, with parking delays reduced by about 40%. It is concluded that the characteristics of vehicle-to-vehicle interactions and the dynamic model of autonomous decision-making behaviors in connected heterogeneous traffic flow can provide a theoretical basis for improving the efficient operation of heterogeneous traffic flow in bottleneck areas.