search

KumarRobotics / conformal_lattice_planner

116 stars 43 forks source link

A question about implemenation of hybrid A* search #4

Closed KailinTong closed 3 years ago

KailinTong commented 3 years ago

Dear Ke,

this part is not really clear in the paper: "To ensure reasonable runtime, hybrid-A* is applied. Longitudinal speed is discretized to three intervals, while the temporal dimension is left undiscretized."

How is the longitudinal speed discretized? what is the temporal dimension?

And could you kindly tell me which part of the code corresponding to the Hybrid A* search?

Many thanks and kind regards, Kailin

ke-sun commented 3 years ago

Just to clarify, spatiotemporal lattice planner is not proposed by this work. For the actual details, it's recommend to follow the work by M. McNaughton, et, al.

An implementation of it in this repository can be found in spatiotemporal_lattice_planner.cpp. As for how speed and time are discretized, please take a look at line 97-118 in spatiotemporal_lattice_planner.h.

KailinTong commented 3 years ago

Thank you for your reply.

So your Feedback Enhanced Motion Planning does not use Hybird A*. Do I understand correctly?

ke-sun commented 3 years ago

You are right. It is not used in the proposed work.

To avoid some potential confusions, I mean it in an informal way of saying that spatiotemporal lattice planner applies Hybrid A*, Since time and velocity dimensions are dicretized, a state with a lower cost dominates others in the same cell even if they are not exactly the same states.