The traffic accident is the prerequisite of the traffic accident reconstruction. In this study, the traffic accident (forward collision) and traffic accident reconstruction (inverse collision) simulations are conducted to improve the quality and accuracy of the traffic accident reconstruction. The vehicle and tire models are used to simulate the trajectories for the post-impact motion of the vehicles after collision. The impact dynamic model applicable to the forward and inverse collision simulations is also provided. The accuracy of impact analysis for the vehicular collision depends on the accuracy of the coefficients of restitution and friction. The neural network is used to estimate these coefficients. The forward and inverse collision simulations for the multi-collisions are conducted. The new method fur the accident reconstruction is proposed to calculate the pre-impact velocities of the vehicles without using the trial and error process which requires the repeated calculations of the initial velocities until the forward collision simulation satisfies with the accident evidences. This method estimates the pre-impact velocities of the vehicles by analyzing the trajectories of the vehicles. The vehicle slides on a road surface not only under the skidding during an emergency braking but also under the steering. A vehicle over steering or cornering with excessive speed loses the traction and leaves tile yaw marks on the road surface. The new critical speed formula based on the vehicle dynamics is proposed to analyze the yaw marks and shows smaller errors than ones of the existing critical speed formula.