Details about sensors and agent format

24 views
Skip to first unread message

DanielFinol

unread,
Dec 14, 2011, 11:37:17 PM12/14/11
to opennero
Hello:

I'd like to have some specifics about the sensors values, concretely:

1.- Enemy sensing radars: What values can they take? What's the
maximum value they can take? (If possible: What function of the input
are they?)

2.- 'Facing enemy' indicator: If I'm 'facing' an enemy (he is within
2 degrees), but there's a wall between us, will this activate? (This
is binary: 1 when facing an enemy, 0 otherwise, right?).

Also, where can I find a detailed description of the format with which
rtNEAT agents are represented in the team file (including which
sensors are which)?

Many thanks in advance.

Igor Karpov

unread,
Dec 14, 2011, 11:51:38 PM12/14/11
to open...@googlegroups.com
On Wed, Dec 14, 2011 at 10:37 PM, DanielFinol <danie...@gmail.com> wrote:
Hello:

I'd like to have some specifics about the sensors values, concretely:

 1.- Enemy sensing radars: What values can they take? What's the
maximum value they can take? (If possible: What function of the input
are they?)

Hi, the sensors are capped at 1, and they start activating at 0.2 (which means there is a single target in range but it is as far as the sensor maximum radius). So, as the target gets closer (or as the sensor starts seeing more targets) the value increases up to 1. You can see the exact way in which this is done by looking at the source of a radar sensor:

 
 2.- 'Facing enemy' indicator: If I'm 'facing' an enemy (he is within
2 degrees), but there's a wall between us, will this activate? (This
is binary: 1 when facing an enemy, 0 otherwise, right?).

Yes, this should stil activate.
 
Also, where can I find a detailed description of the format with which
rtNEAT agents are represented in the team file (including which
sensors are which)?

The sensors are added in a sequence:


First there are the WALL_SENSORS, followed by the FLAG_RADAR_SENSORS, followed by the ENEMY_RADAR_SENSORS. Then we have two sensors for friends (center of mass) and one more for targeting.  Finally there is a bias input which is just a constant - it is needed to provide the network with input energy even when input sensors are 0.

 
Many thanks in advance.

DanielFinol

unread,
Dec 15, 2011, 12:13:44 AM12/15/11
to opennero
Thanx!

On Dec 15, 12:51 am, Igor Karpov <ikar...@cs.utexas.edu> wrote:


> On Wed, Dec 14, 2011 at 10:37 PM, DanielFinol <danielfi...@gmail.com> wrote:
> > Hello:
>
> > I'd like to have some specifics about the sensors values, concretely:
>
> >  1.- Enemy sensing radars: What values can they take? What's the
> > maximum value they can take? (If possible: What function of the input
> > are they?)
>
> Hi, the sensors are capped at 1, and they start activating at 0.2 (which
> means there is a single target in range but it is as far as the sensor
> maximum radius). So, as the target gets closer (or as the sensor starts
> seeing more targets) the value increases up to 1. You can see the exact way
> in which this is done by looking at the source of a radar sensor:
>

> http://code.google.com/p/opennero/source/browse/trunk/source/ai/senso...


>
> >  2.- 'Facing enemy' indicator: If I'm 'facing' an enemy (he is within
> > 2 degrees), but there's a wall between us, will this activate? (This
> > is binary: 1 when facing an enemy, 0 otherwise, right?).
>
> Yes, this should stil activate.
>
> > Also, where can I find a detailed description of the format with which
> > rtNEAT agents are represented in the team file (including which
> > sensors are which)?
>
> The sensors are added in a sequence:
>

> http://code.google.com/p/opennero/source/browse/trunk/mods/NERO/NeroE...http://code.google.com/p/opennero/source/browse/trunk/mods/NERO/const...

Reply all
Reply to author
Forward
0 new messages