Near-optimal trajectory generation, using a compound B-spline interpolation and minimum distance criterion with dynamical feasibility correction

Trajectory generation for robotic vehicles tries to provide real time computing of a collision free path from one point to another, which usually involves minimizing a performance measure, to accomplish a desired task. A novel hierarchical method for generating near-optimal and collision free trajectories is proposed here to be used on-line in three-dimensional space. At the top level, the off-line optimal trajectory problem is solved in a complex environment, using B-spline interpolation and genetic algorithm, while considering the dynamic constraints of the vehicle. At the intermediate level, the path is modified for any possible on-line intersections with unexpected obstacles, using a minimum distance correction technique. The feasibility of the generated path is assessed at the lowest level by considering dynamic constraints of the vehicle. A novel method is presented at this level for correcting the path and obtaining near-optimal and dynamically feasible trajectories. Each part is assessed by a separate robotic vehicle to ensure the capability and performance. Lastly, the consistency between the levels of hierarchy is evaluated by presenting a differential-drive robot example. The final path is also verified with two commercial software codes for path planning and optimal trajectory generation.