Virtual track train (VTT) is a new urban public transportation system that adopts rubber wheel and rail transit management mode. Its characteristics of all-wheel-driving all-axle-steering (AWDAAS) enable this multi-articulated vehicle to run along the target path (virtual track) with high precision. This paper proposes a new traction guidance cooperative control strategy aiming at path-tracking and hinge forces optimisation. Firstly, the vehicle dynamics model is established using the two-step method, which mainly includes the centre of gravity (CG) forces model between the wheel steering angles, torques and the generalised forces at every unit. Secondly, the vehicle multi-body dynamics model is built with the CG generalised forces as the control input and carries on the concrete modelling for the three-unit six-axle AWDAAS-VTT. This modelling method can satisfy the rapid expansion and reconstruction according to the different marshalling numbers and axles. Then the hinge forces are analysed and expressed quantitatively by vehicle state and control inputs. A traction guidance cooperative control strategy based on a hierarchical framework is proposed, which has mainly three parts: the cooperative distribution of path-tracking and longitudinal speed targets, the CG generalised forces calculation of every unit based on improved model predictive control (MPC) and Stanley algorithm, and the control allocation (CA) of wheel torque and angle. The control strategy is verified by Co-simulation. The results show that the proposed control strategy can achieve excellent tracking accuracy and effectively reduce hinge forces. In addition, the vehicle under cooperative control also has excellent curve passing stability and continuous curve adaptability. The control strategy has high robustness under different running speeds, curve radii and payloads. Even under the extreme condition of passing through the R20 m curve at 14 m/s, the maximum tracking error is less than 0.035 m.