Figure 1 depicts the process overflow of the long-term prediction algorithm. The different steps of the long-term prediction algorithm is explained here. The process of the long-term prediction algorithm can be separated in 5 parts. 1) For each vehicle, PRIDE retrieves its current position, current velocity and the number of lanes of the portion of the road where the vehicle is situated. 2) For each lane, the long-term prediction algorithm computes each possible sequence of actions that a vehicle could perform. 3) For each action sequence, a cost for performing this action sequence is computed and a predicted vehicle trajectory is computed. A predicted vehicle trajectory is a vector whose origin is the current position of the vehicle and whose extremity is the predicted position the vehicle plans to reach. 4) For each predicted vehicle trajectory, a cost for being over or under the speed limit is applied, a cost of collision is applied and the final cost of the predicted vehicle trajectory is recomputed. The cost of each predicted vehicle trajectory is converted to a probability. The probability is inversely proportional to the cost, the higher the cost, the lower the probability. 5) The predicted vehicle trajectory with the highest probability is used by the vehicle.

Return to PRIDE page