Connected and autonomous vehicles (CAV) have seen a rapid development over the past decade. However, wide diffusion of high level CAVs is still decades to come, and will depend on many technological, policy and public acceptance factors. Merging a traditional Bass Diffusion Model with a discrete choice model in a system dynamics approach, this study modelled CAV diffusion from 2020 to 2070 in the UK, considering mode choices of CAV private car, CAV car/ride sharing and CAV bus, their possible impacts on road network performance and sus-tainability, and the feedback of these impacts to CAV diffusion. Results of this study suggest that without in-terventions CAV diffusion will start to increase rapidly from 2035, and reach market saturation of 98% in around 2057. CAV diffusion will lead to reductions in average travel time, average travel cost, carbon emission and traffic accident. Training campaign, which prepares the general public to be ready for CAVs, is more effective in accelerating CAV diffusion than marketing campaign, which mainly targets the innovators and early adopters. Promoting shared CAVs and CAV public transport can contribute to more sustainable and more affordable mobility with CAVs, although this may lead to smaller CAV market size in terms of CAV sale, and the market size may reduce at a higher rate than sustainability enhancement.