Dynamic window approach

Dynamic Window Approach ( DWA short, engl. " Dynamic window method " ) is a strategy for collision avoidance with robots, which was published in 1997 by Dieter Fox, Wolfram Burgard, and Sebastian Thrun. The strategy takes on other than a general method in view of the dynamics of the robot, and in particular a limited speed and acceleration. She goes for it from all sorts of rules and tries to find in this quantity, the optimal control. The aim is to motion planning for robots with response to dynamic environments and unforeseen events at runtime.

With DWA, the movement of the robot is planned in advance only for small time intervals. In this case, the kinematics and dynamics of the robot is taken into consideration, in particular the limits of speed and acceleration. The search space for possible arrangements of the robot is to be limited to those provided by these limits at any time depending allow a safe stopping the robot. In addition, the search space is limited to speeds that can be reached in the pre-planned time interval. Speeds corresponding to these constraints, form the eponymous dynamic window, the dynamic window.

On the possible translational and rotational speeds of the dynamic window, a power functional is maximized. The quality function in the original publication contains as weighted shares

  • Approach to the target point
  • Forward speed
  • Distance of the trajectory to the next obstacle

Thus, the order determined control ensures avoiding collisions with obstacles for a rapid approach to the target point. In the publication, results are presented on various robotic platforms in different environments, leading to a robust obstacle avoidance at speeds up.