Post by andrew on Apr 2, 2013 5:22:22 GMT -5
Hi there
I have been thinking a lot recently about how to improve the vision channel.
currently the agents can easily recognize an opponent agent and follow it if in view, but when the crowd gets too big, they are having a lot of problems deciding which agent to follow.
In human logic we all set a goal, to get from point A to point B
And the same logic happens when we are deciding which person we want to follow in a crowd.
If there are three different people and i want to follow one of them, then i will have to make a decision. obviously one would prefer to follow the closest person, but what do you do if they are all equally close to you ?
you simply just choose one of them, because you will have a goal that says follow one of three persons. and so you make that one person your target.
Okay so to clarify my theory in the use of Miarmy.
there are 6 agents called agent A. and 12 agents called agent B
Agent A is told to follow agent B, we can see that there is way too many agent B in order for Agent A to decide which ones they should follow.
but if we follow my theory, Agent A will get some variables to chose from.
variable 1: follow closest to you. this will eliminate some of the agents from being followed thus making it easier for Agent A to decide.
variable 2: chose one of xx agents to follow. this is simply just a random variable that says chose a number between 1 and xx.
the most important variable now, is the one that tells A agent to keep following Agent B, even though we can't see him.
variable 3: if Agent B is out of sight, follow last known position.
this variable is to make sure that we stay on our goal, remember we still want to go from point A to point B.
another variable that perhaps should have been number one or two. Is a tell and hear variable.
when Agent A chooses a target, all corresponding Agents will respond to it and choose a different target.
i don't know for sure if this variable should even be here because now when i think about it, it would be great to have multiple agents following the same thing.
anyway this is just some thoughts i have made, i hope this could be useful for your next update.
ps. i was wondering if you could extend the aiming channel to work with other agents instead of just geometry?
regards.
andrew
I have been thinking a lot recently about how to improve the vision channel.
currently the agents can easily recognize an opponent agent and follow it if in view, but when the crowd gets too big, they are having a lot of problems deciding which agent to follow.
In human logic we all set a goal, to get from point A to point B
And the same logic happens when we are deciding which person we want to follow in a crowd.
If there are three different people and i want to follow one of them, then i will have to make a decision. obviously one would prefer to follow the closest person, but what do you do if they are all equally close to you ?
you simply just choose one of them, because you will have a goal that says follow one of three persons. and so you make that one person your target.
Okay so to clarify my theory in the use of Miarmy.
there are 6 agents called agent A. and 12 agents called agent B
Agent A is told to follow agent B, we can see that there is way too many agent B in order for Agent A to decide which ones they should follow.
but if we follow my theory, Agent A will get some variables to chose from.
variable 1: follow closest to you. this will eliminate some of the agents from being followed thus making it easier for Agent A to decide.
variable 2: chose one of xx agents to follow. this is simply just a random variable that says chose a number between 1 and xx.
the most important variable now, is the one that tells A agent to keep following Agent B, even though we can't see him.
variable 3: if Agent B is out of sight, follow last known position.
this variable is to make sure that we stay on our goal, remember we still want to go from point A to point B.
another variable that perhaps should have been number one or two. Is a tell and hear variable.
when Agent A chooses a target, all corresponding Agents will respond to it and choose a different target.
i don't know for sure if this variable should even be here because now when i think about it, it would be great to have multiple agents following the same thing.
anyway this is just some thoughts i have made, i hope this could be useful for your next update.
ps. i was wondering if you could extend the aiming channel to work with other agents instead of just geometry?
regards.
andrew