Real-time trajectory optimization treating robot paths as elastic bands subjected to virtual obstacle forces and kinematic vehicle bounds.
The TEB planner optimizes path geometry and execution time simultaneously using non-linear least squares (g2o solver). Cost functions incorporate maximum linear and angular velocity, wheel acceleration limits, and obstacle clearances.
By estimating the velocity vectors of moving pedestrians and forklifts, the planner predicts collision cones, initiating proactive passing maneuvers before the robot enters the human safety zone.