A human-machine co-driving framework considering driving intention was proposed to address the human-machine conflicts caused by the goal misalignment between natural drivers and automated systems in human-machine co-driving process. The framework integrated a vehicle kinematic model with the driver’s steering commands to generate short-term predictive trajectories that represented driver intention. The road environment was discretized into free units through the Delaunay triangulation algorithm and a heuristic graph search algorithm was employed to generate the desired feasible regions online. By incorporating feasible region boundaries and vehicle stability constraints into the model predictive control (MPC)-based optimization problem, a shared controller was designed with the objective of minimizing human-machine control deviation, which maximized the driver’s operational freedom. Simulation results demonstrated that the proposed planning and control strategy not only exceled in ensuring driving safety but also fully considered the driver’s behavioral intention, significantly enhancing the driver’s acceptance of the driving assistance systems. The proposed human-machine co-driving framework achieved an integrated coupling of planning and control at the algorithmic level, offering a novel approach for realizing efficient human-machine cooperation based on driving intention.