| |
In lugnet.robotics.rcx.robolab, Mike Reddy writes:
> HI,
>
> I am building a robot ant and using a light sensor to detect lines to
> follow. However, the algorithm I am using must not be deterministic; i.e. I
> need to do a sensor sweep and build a graph of the relative light levels at
> the left, front and right of the ant, then to use a weighted probability to
> choose the direction - very like real ants with pheremone trails. Is it
> possible to do a data log for the light sensor, then have the robot itself
> do the calculation to determine the direction, rather than uploading and
> post processing with Investigator? I am using a mac and robolab 2.5.
>
> Ideally, I want the robot movement to be autonomous, as I will be using the
> IR channel for other communications with different ants, food, nest etc.
>
> Thanks
Hi Mike,
Unfortunately there is no possibility under Robolab 2.5 that the RCX might
itself access the data collected in the log. Robolab 2.5 is based on
LEGO-firmware 2.0, which has many new features, but not this one.
If you want to do your analysis in real time without IR-communication to the
PC, you must deal with the RCX limits, which are:
32 global variables and 16 local variables per task (maximum 10 tasks).
You could for instance use a classical data-reduction technic, which is the
least square polynomial. (We explained the technic at
http://www.convict.lu/Jeunes\Math\least_squares_polynomial.htm .)
You could collect the data in function of the angle of the sweeping
light-sensor - I suppose you use some kind of angle-measurement to know
where the sensor points to -, then compute the different sums. I think a
second degree polynomial function would do the job.
Of course this is again tricky in Robolab with the integer-number limits,
but not impossible.
After all, that kind of problem invites you to change to leJOS!
Claude
|
|
Message is in Reply To:
 | | Robot Ant and Data logging
|
| HI, I am building a robot ant and using a light sensor to detect lines to follow. However, the algorithm I am using must not be deterministic; i.e. I need to do a sensor sweep and build a graph of the relative light levels at the left, front and (...) (25 years ago, 4-Apr-02, to lugnet.robotics.rcx.robolab)
|
2 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
