Date: Fri, 30 Dec 2005 14:12:11 -0600 (CST) Subject: radios ok ; driving around with gps ; stanford's touareg X-UID: 116 Content-Type: IMAGE/JPEG; NAME="img2490.jpg" Content-Type: IMAGE/JPEG; NAME="img2496.jpg" Content-Type: IMAGE/JPEG; NAME="img2498.jpg" Content-Type: IMAGE/JPEG; NAME="img2499.jpg" Content-Type: IMAGE/JPEG; NAME="img2505.jpg" Both radios are in Penguin peppermint tins now. I'm waiting another day or two before handling them as some silicone around the pigtail cable needs to cure. I tested them before applying the silicone and they seem 100%. The GPS receivers also work. I drove the strip of White Rock Lake park near my apartment and also a few miles down Buckner and Garland roads to test them. The Garmin 18 LVC (low voltage CMOS) is an industrial OEM product so is rugged and has a magnetic mount built in. This is convenient as they stick to the top of my car and don't have any problems with harsher environmental conditions (immersible in 1 meter of water for 30 minutes, operational temperature range of -30 C to +80 C). Unfortunately, the Dell D610 is marginal. The LCD malfunctions when it is too hot. My guess is that the liquid crystals don't polarize properly when they are too hot. Tipping the display down for a moment clears up the problem. This indicates that the thermal problem is due to sunlight on the front of the display, not the magnesium plate on the back. The only real solution is running the AC to keep the computer cool. I'm starting to understand why military grade rugged laptops are necessary for field use. Consumer grade equipment is designed to operate indoors in a sheltered and climate controlled environment. Industrial and military grade equipment must be designed differently to work in extreme environments (and outside my parking lot on a sunny winter day is "extreme" enough to make this laptop not work - this is not Iraq). That idea I had to steal USB port power from the laptop for the GPS receiver and 900 Mhz radio is impractical. Had I naively tried this, I could have burned something out. Devices requiring more than 100 milliamps (the 900 MHz radio) must request additional power from the host (the laptop). There are also fairly tight electrical requirements such as bypass capacitance on the device. This means that there is more engineering than you might think on even simple devices like USB powered cup warmers. That cup warmer must identify itself over the USB bus and request more power from the computer it is plugged into. Everything is intelligent now. So I'm using a pack of four AA NiMH cells. While the nominal voltages are specified at 1.2 volts, when freshly charged the voltage is 1.375 volts. A pack of four cells is 5.5 volts instead of 4.8. That's somewhat higher than I anticipated. Fortunately, the GPS receivers have a range of 4 to 5.5 volts and there's a little resistance in the wire and connectors from the battery pack to the devices. So I was just inside the electrical specification. However, the radios require 4.75 to 5.25 volts. So had I tried plugging them into the battery pack, they might have been damaged. I'll use a switching diode to drop the voltage from 5.5 down to 5 volts. I read on Wired about Stanford's modified Touareg that won the Grand Challenge. There is some interesting history and conclusions. The Germans were the first to create a self driving car. They did this years ago with Pentium class technology. That's roughly equivalent to what I have on my robot. The Geode processor is a Pentium with MMX extensions. The difference is that I have about 1/1000 the amount of money sunk into my project. Instead of tens to hundreds of thousands of dollars for laser rangefinders, cameras, etc, my robot will have less than $150 of webcams, IR distance measuring sensors and two laser line levels from Walmart. I'm curious if it can exhibit good performance despite being so cheap. More interesting is the probabilistic / heuristic reasoning involved. Stanford does claim they had to have probabilistic reasoning in their vehicle. My thought after the first Grand Challenge is that many of the robots could have completed the course if they were self aware. Failure modes including become stuck on a rock and then going to full throttle to try to break away. A self aware machine would realize that if it is applying full throttle and not moving, it will destroy itself. Or a self aware machine, like a person, understands the concept of illusions. So if you are driving down a road you know to be clear but for a second see a big rock appear ahead, you can debate whether the rock is real or rather some problem with your eyes and brain. In that case, you might ignore the rock as it is not really there. Unfortunately, while Carnegie Mellon was very open about their approach for the first race, Stanford is incredibly secretive. They reveal nothing technical. What they do say confirms what I had suspected which is enough to encourage the approach I had anticipated.