In this paper, a collision avoidance planning and control strategy is proposed for an unmanned ground vehicle (UGV) at intersections based on digital twin (DT). A DT system for the UGV is established with four layers, i.e., physical space, virtual space, application service and data process. In the virtual space, collision avoidance trajectory planning is achieved by combining an A-star algorithm with quadratic programming based on real-time environment and obstacle vehicle information. Taking a planning trajectory as a desired trajectory, a bidirectional tracking control method is devised to realize synchronous tracking for a virtual UGV and a physical UGV. A nonlinear extended state observer, a kinematic controller and a finite-time sliding mode controller are designed in the bidirectional tracking control method. Experiment results indicate effectiveness and superiority of the collision avoidance planning and control strategy for the UGV based on the DT system.