article

When I said they weren't perfect, I was speaking of the models themselves, not the Agent system or AGV system. There is nothing in the Agent system or AGV systems that needs to be fixed. What I was pointing to that wasn't perfect was the heuristic that determined whether an object was "in front of" an AGV. If an object gets into proximity with an AGV, and it is "in front of" the AGV, then the AGV should stop. If an object is "behind" the AGV, then the AGV should continue so that it will get out of proximity. I was doing some geometric math that held up most of the time, but not all the time, especially when they were on a turn. So, if you want to "fix" this, then you should come up with a better heuristic for determining when another AGV was "in front" of a given AGV, or in other words, better logic to determine whether an AGV should actually stop vs. continue when it comes into proximity with another object. And that heuristic may be dependent on the specific situation that you are simulating, or based on the actually hardware (sensors, etc.) that will be installed in real life. So I don't know if I could give you a "catch all" heuristic.