System Design Overview
The last chapter ended with some discussion around selecting a motor unit. Following some advice from a friend, I made some changes to the figures for the calculator. Updated for the following:
Best I could find for rolling/sliding coefficient of friction metal on concrete was 0.03.
80% system efficiency
10* slope
90ft/min ~ 350ft/min
New Results
Required Torque : 3560oz/in = 25Nm
Accel Torque : 544oz/in = 4Nm
Load Torque : 1829oz/in = 13Nm
New figures sound much more reasonable/manageable. I think at this point I can just pick a motor and play around with gear ratios until I hit a torque / RPM combination that fits the bill, purchase all the components going into the head, hook them up to a breadboard and get everything communicating to start some tests. Maybe even hook it all up to an RC car as a proof of concept?
System Design:
You only get distance to the user from a BT connection, so you don't know angle to the target. In order to locate the user, we keep track of previous positions to triangulate their position. To keep track of obstacles we can use the ultrasonic distance sensor, and have the head swivel back and forth to create an echolocation cone. My initial idea was to feed all these parameters into a singular neural network, the output being commands to give the two motors. Three obstacle detection beam measurements, one BT distance measurement, as well as signals sent to the two motors is 6 parameters to keep track of. 12 total parameters required for the simplest form of user location if we want to keep minimum robot movement history, 18 total parameters if we want to remember two previous positions.
Started looking into neural networks being run on Arduino, and came across a good article here with some example code for building Artificial Neural Networks (ANNs) to work with Arduino. http://robotics.hobbizine.com/arduinoann.html Unfortunately, the Arduino has a pretty limited SRAM, so the networks are limited to a size about 7x8x4, half of what our parameters are asking for.
I've found what looks like to be a suitable auxiliary SRAM chip, which would add 256Kb of room to work with. The chip operates at 20MHz at 3V, which I'm hoping will be enough to keep the system relatively responsive. Libraries exist which allow for communication over SPI, and SPI supports up to 4 peripheral devices, allowing me to continue use of SPI communication for MicroSD as well as SRAM expansion.
In the next chapter, we’ll receive a delivery of parts and begin work on estimating distance over Bluetooth using RSSI.
