A Self-Learning Neural Network Robot with multiple BASIC Stamp Controllers?
demosthenes
Posts: 2
Alright, lets see if I can explain this coherently...
I am trying to build a 4 to 6 legged walking robot with motion and nervous system based on biological entities (most notably the cockroach) I'm going about this with a concept of external cellular neural networks (which control the legs, monitor sensor input, and general threshhold limits, and communicate with yhe central core (a BASIC Stamp 2 IC of some sort) over one or two I/O lanes. The cells could themselves be BASIC Stamps, but I'd be unwilling to put anything above a BASIC Stamp 1 to these tasks because they dont require that much computing power. The BASIC Stamp 2 will act as a brain and maybe a coordination outfit for the whole operation, while the cells themselves take care of their own functions autonomously with no overrides from the "brain"
The Cell (if for example applied to one of my robot's legs) would recieve a signal from the BASIC Stamp 2 brain to "begin" over I/O line 0. It would respond with a constant signal (over the same I/O line) to let the brain know that it is indeed busy and functioning properly if there is an error for any reason, the leg will stop and a program halt will occur in the main stamp. While the signal is being sent over I/O 0, the BASIC Stamp 1 (at the core of the cell) will be signaling servos to start and stop, based on the amount of degrees that the servos move. The leg will lift up, move forward, and set down, requiring two servos (and 4 I/O lines?) once the movement of the leg is finished, the BASIC Stamp would halt the constant signal and let the BASIC Stamp brain know that the sequence has been completed. The BASIC Stamp "brain" would monitor and organize the progress of all of the legs and input from the sensors, and run a learning program designed to mimic a neural network so that it could learn and respond to new situations as it comes accross them (with little to no input of new data from me)
My questions are these:
Does the BASIC Stamp 2 IC have enough power to do all this?
Is the idea feasible in concept? (I think it is, but thats just me)
I am trying to build a 4 to 6 legged walking robot with motion and nervous system based on biological entities (most notably the cockroach) I'm going about this with a concept of external cellular neural networks (which control the legs, monitor sensor input, and general threshhold limits, and communicate with yhe central core (a BASIC Stamp 2 IC of some sort) over one or two I/O lanes. The cells could themselves be BASIC Stamps, but I'd be unwilling to put anything above a BASIC Stamp 1 to these tasks because they dont require that much computing power. The BASIC Stamp 2 will act as a brain and maybe a coordination outfit for the whole operation, while the cells themselves take care of their own functions autonomously with no overrides from the "brain"
The Cell (if for example applied to one of my robot's legs) would recieve a signal from the BASIC Stamp 2 brain to "begin" over I/O line 0. It would respond with a constant signal (over the same I/O line) to let the brain know that it is indeed busy and functioning properly if there is an error for any reason, the leg will stop and a program halt will occur in the main stamp. While the signal is being sent over I/O 0, the BASIC Stamp 1 (at the core of the cell) will be signaling servos to start and stop, based on the amount of degrees that the servos move. The leg will lift up, move forward, and set down, requiring two servos (and 4 I/O lines?) once the movement of the leg is finished, the BASIC Stamp would halt the constant signal and let the BASIC Stamp brain know that the sequence has been completed. The BASIC Stamp "brain" would monitor and organize the progress of all of the legs and input from the sensors, and run a learning program designed to mimic a neural network so that it could learn and respond to new situations as it comes accross them (with little to no input of new data from me)
My questions are these:
Does the BASIC Stamp 2 IC have enough power to do all this?
Is the idea feasible in concept? (I think it is, but thats just me)
Comments
Sorry to say but this is not a neural network, since your 'cells' do not treat the information in itself before propagating the signal to the next cell.....
What you are describing is a master/slave communication...
Well, yes you can, the 'brain' will handle the sensors and logic, while the cells (on per leg) will handle the low level servo driver.
There are no problems to use such an arcitecture wich BS ICs, they have enough power.
++
The neural network will be a virtual neural network running in the BS2. The BS1's will be slaves but will the original BS2 be enough for all the I/O lines I need, or should I get the extended I/O Version? Is the original BS2 fast enough for my purposes, and does it have enough space for the type of program I have in mind, or do I need to go for one of the other models?
The cells themselves will be acting like reflexes. You tap the knee, it jumps a bit. In the same way, you send the pulse, the leg moves on its own.
Sending this at 6 in the morning :P
Demosthenes
how many inputsand outputs you will have
how much data
how fast you need to send the data
how much processing of the data you need to do
realtime or can it be buffered
then, think about how to compress the data
or multiplex signals to reduce data rates and IO pins
Pretty tough to know if the BS2 can work without the above questions answered first.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Think outside the BOX!
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
·1+1=10