Smac Planner Hybrid Issue about navigation2 HOT 3 CLOSED

Yes it is present in the global costmap, and its not lethal, as you said higher cost but shouldn't Smac Hybrid planner plan around it based on the cost penalty with conjunction of global costmap ?

Not necessarily, there are trade offs beyond just cost, like shortest length. We couldn't want a robot necessarily to add 100m to a path to stay in 0 cost space over entering 1 cell that is 1 cost, right? So we need to decide on a trade off between minimum length and reasonable cost to travel. We do that with the traversal function of cost_to_travel = cell_length * (1.0 + cost / 254 * weight). That makes us balance the cost to travel by both the length of the primitive with the normalized, weighted cost. A higher cost weight would make traveling in higher cost spaces more expensive. Figure 5 of this paper highlights how the change in cost penalty changes behavior.

Looking at your example, it seems like the extra ~+2m to go around it isn't proportionate to the extra cost from such a small obstacle (its only 1-2 pixels thick). I might expect a different result if your virtual obstacle there was filled. What you show is more or less expected and good behavior. If you wanted that are to be not traversed at all, you should mark it with lethal cost via whatever custom mechanism you've added that with. Though at least from your snippet, it looks like possibly just a user error that it really is lethal space but you have your obstacle_layer after your inflation_layer in your configs. Those are ordered; make sure your inflation layer is last. That very well may resolve your concern.

Note: In general, please don't post pictures of parameters, include them as code snippets for the sections that are important.

from navigation2.

Its unclear from your diagram: is that present in the global costmap used for planning or the local costmap used for control, or both?

That's doesn't look like a lethal obstacle to me (yeah?), that's just a higher cost region as evidenced that there is no inflation around it. The cost functions weight distance with cost, so in this situation, it was better to go through 1-2 cells of higher cost than to diverge dozens. Thus, this looks fine to me.

from navigation2.

Yes it is present in the global costmap, and its not lethal, as you said higher cost but shouldn't Smac Hybrid planner plan around it based on the cost penalty with conjunction of global costmap ?
This is my below global and local costmap setup so far,

from navigation2.