Worked a whole whole lot on the software. Finished.

Worked a whole lot on the software.

I got a cigarette plug socket for $10. I also made two relay control circuits. Worked a whole lot on the software

I made a new post on the web log at mobilerobotics.com. Worked a whole lot on the software.

Today I ordered a 2ah 12V sealed lead-acid battery and four 5V DPDT relays that are rated at one amp. This is all I need to make the test platform work. Later I designed a low battery warning circuit so that the computer could shut down correctly before the power inverter shut down. I tested it, and it worked!

I tested the Robotix motors for the platform on 9V DC. They weren’t strong enough. Later I tested them on 12V DC, and they worked well.

I successfully taught a net to tell the difference between red, green, and yellow blocks of the same size. I also started a test platform. It uses Robotix motors and has the Mini-Itx on it. I plugged it into the power inverter, and plugged the two cameras into it. It needs relays to control the motors, and batteries. The battery voltage has to be 12V. I could use ten AA batteries, but it would only last an hour. A rechargeable sealed lead-acid (sla) battery would be more convenient.

I added a counter to my website. I also made a post on the blog at www.mobilerobotics.org.

I started a website at http://dnps-brad.home.comcast.net

I changed the software design by adding a low battery detector. This would just shut down the computer and stop the vehicle. Here’s the new software design:

I also made it so that the network’s settings can be accessed from the network interface program, and then moved the network settings form from the neural network program to the network interface program.

I also made it so that while the network is learning, the network interface program shows which picture it’s “studying”.

I made it so the neural network stops learning when the percent correct is above a certain amount that can be chosen by the user, instead of always being 100%.

I got a 150W 12V DC to 110V AC power inverter. I also worked on the neural network’s interface a lot.

I made the neural network software a DLL and made plans for how the software will work. A lot of changes must be made. The ANN software will not have a video portal, but will get information from a different program. This is so that only one video portal has to be made even if there are going to be multiple neural networks. This will speed everything up. Besides the ANN program, there will be another program that makes data-packages and gives them to neural networks, and then tells them to learn. A third program will be connected to neural networks (that were taught using the second program) that are processing live camera images. This program will use the networks’ outputs to do something.

I also made a major change to how the ANN’s learning process. Instead of learning until the outputs are close enough to the correct outputs and then moving on to the next picture, it just changes a certain percent of all the weights and then moves on. Right now the percent is set to 0.04%. The purpose of this change is to make it so that if there is a data-video in which a few frames have incorrect information, the network will not spend a very long time getting wrong information. Instead, it just moves on.

I installed Visual Studio 6.0 on Robby, the Via computer, and then MSDN Libraries, a service pack, and driver utilities.

I build a PC out of the Mini-Itx and parts of my old computer. I then installed Windows 2000 server operating system.