Here is an issue I guess everyone's having but I haven't figured out yet a satisfying (simple, reliable) way of solving it. It involves an agent and its target in a room. The target is on an unreachable area (not on the navmesh). In the first image, the agent goes towards its target because it can reach it (its stopping distance is its attach reach distance). But in the second picture the target has moved away and the agent can't reach it anymore. So the agent decides to get out of the room to try and attack its target from the other side of the wall. Because in computer-logic the distance would be technically shorter. Of course it doesn't make any practical sense. So is there a way to keep this behavior from happening? In other words is there a concept of "walls" in the navmesh system?