Agent Movement

[thomasGodfrey]

• [x] Currently have significant performance hits in cases where an agent is moving towards another moving agent. This is because the follower will re-evaluate their path finding every simulation tick because their target is moving. Need to change this so that the agent moves to the 'last known location' then re-evaluates if the target is no longer there
• [ ] Pathfinding is too strict - agents follow the quickest path which often involves hugging walls. Could be countered by adding more weight to spaces around walls to discourage paths too close to walls

[szschaler]

Another alternative that I have seen in other ABMs is to define paths explicitly. I think I saw a video from Slingshot, for example, where they allowed such paths to be annotated graphically on top of building plans.

[thomasGodfrey]

Considering pre-computing paths from each location:location pair that exists in the map. Yet to see if this would cause the model loading to become prohibitively slow