I've done that something like that.[1] It's appropriate where there's a significant cost to detecting obstacles, because it tests few unnecessary cells.
It heads to the goal until an obstacle is reached, then follows the wall. Unusually, it forks and follows both the left and right wall simultaneously. It's not always optimal, but the optimal algorithms such as A* have to test more cells.
It heads to the goal until an obstacle is reached, then follows the wall. Unusually, it forks and follows both the left and right wall simultaneously. It's not always optimal, but the optimal algorithms such as A* have to test more cells.
This algorithm runs my NPCs in Second Life.
[1] https://github.com/John-Nagle/lslutils/blob/master/npc/obsol...