In order to travel in a chaotic and dynamic environment, an autonomous vehicle requires a motion plan. This motion plan ensures collision free, optimum travel without violating any traffic rules. The optimum solution for path planning problem exists in higher dimensions, however, with the help of useful heuristics the problem can be solved in real time, which is required for real time operation of an autonomous vehicle.
There are different well established techniques available to plan a collision free kinematically traversable path. One of such techniques is called conformal Lattice planner. However, the legacy version of conformal lattice planner is not optimized and also is prone to fail under specific dynamic environment conditions. Moreover, the legacy version of conformal lattice planner is also not road aware. Due to this reason it is a semi optimized way to solve the motion planning problem.
This paper discusses a modified version of conformal lattice planner, which uses the road information to better plan the paths by filtering out those paths that are going to end up in a collision. This way, the paths that are obviously going to end up in a collision are pre-eliminated and they do not enter the computationally heavy path-optimization and collision checking cycle. Also the modified conformal lattice planner makes sure that the planned motion paths are more prone to guide the vehicle to a safer trajectory and at the same time following the traffic rules.
Lastly, paper discusses the inherent problems with the legacy version of conformal lattice planner technique and how the modified lattice planner version is able to eliminate these problems.