In order to effectively solve the problem of poor estimation accuracy of the existing road peak adhesion coefficient estimation method on the non-structural road surface, a road peak adhesion coefficient estimation method based on the extended Kalman filter under the non-structural road surface is proposed. Acquiring vehicle motion response signals such as a vehicle pose, a wheel rotation angle and the like through a common vehicle-mounted sensor, introducing an equivalent suspension model into a traditional vehicle model to calculate motion parameters such as a wheel vertical load and the like, taking the parameters as input, and calculating a coefficient matrix of an extended Kalman filter through a Dugoff tire model; An equivalent suspension model is introduced to optimize the calculation of the vertical load of the wheel, and the acceleration of the vehicle is corrected by combining the position and posture data of the vehicle, so that the identification accuracy of the road adhesion coefficient under the unstructured road surface is improved. The simulation results of multiple driving conditions in Carsim with straight line braking show that the estimation accuracy of adhesion coefficient is improved by more than 6.6%, which proves the effectiveness and accuracy of the road adhesion coefficient estimation method.