<p>As the scale of the Internet of Vehicles expands and the penetration rate of electric vehicles increases, the dynamics, heterogeneity, and multi-objective optimization problems of resource scheduling have become increasingly prominent. The study suggests an autonomous collaborative resource control system for electric vehicles based on the intelligent sensing Internet of Vehicles in order to address this issue. First of all, this research designs a digital twin-driven state modeling and prediction module to complete the accurate depiction and forward-looking deduction of the global state through the dynamic interaction of physical entities and virtual mapping. Then, the study introduces the multi-agent deep deterministic policy gradient algorithm to build a collaborative decision-making mechanism of centralized training and distributed execution. This enables dynamically optimized allocation of charging power, task offloading and communication resources. The simulation test results revealed that the average task cost of the proposed system was only 16 when the task size was 2.0 to 2.5 in a static scenario, and the transmission failure rate of the proposed system was 0.1 when the vehicle speed was 25&#xa0;m/s in a dynamic scenario. The application test results of the autonomous collaborative resource control system for electric vehicles showed that the energy consumption of the charging station of the proposed system under the bandwidth of 5&#xa0;MHz was 145&#xa0;kJ, and the energy consumption of electric vehicles was 81&#xa0;kJ. In summary, the system proposed in this study is a deep integration of intelligent sensing, digital twins, and multi-agent reinforcement learning. This can effectively solve the collaborative bottleneck and dynamic adaptation defects of the traditional control model.</p>