Planning of Safe Trajectories in Dynamic Multi-Object Traffic-Scenarios

Active safety systems have a huge potential of increasing road traffic safety and are a necessary component of vehicles for autonomous driving. An algorithmic challenge of such systems is the prediction of collision-free trajectories in complex scenarios with multiple dynamic objects and curved roads. This work proposes an algorithm able to find safe trajectories in such an environment. The algorithm is dividing the prediction horizon into two intervals. In the first interval multiple collision-free trajectories are computed using an extension of the CL-RRT algorithm, introduced as Augmented CLRRT, which contains a controller to consider the nonlinear dynamics of vehicles. The safety levels of the found trajectories are evaluated using the second prediction time interval, by taking into account the steering effort that is necessary to travel to a safe region on the road. The efficiency of the proposed algorithm is demonstrated exemplarily using a simulation with multiple dynamic objects in a traffic scenario with a curved road.