In order to improve the accuracy of predicting vehicle collision accidents and reduce the probability of collision accidents,vehicle anti-collision algorithm（VACA）based on vehicle to vehicle communication was proposed.According to Beidou location coordinates of vehicle itself and other vehicles in the vicinity,VACA used Kalman prediction algorithm to predict next vertical and horizontal locations of vehicles.VACA used data prejudge method to directly judge whether collision will happen.When it could't be directly judged,safe distance with variable acceleration was calculated,and collision state prediction model was established and solved,then minimum collision prediction time could be obtained.If the minimum time was less than threshold value,alarm signal would be sent to remind driver.Simulation results show that compared with IVCWM and ECWA,VACA can predict collision accident more accurately and has certain application value.