Date: Wed, 28 Sep 2005 01:50:43 -0500 (CDT) Subject: robot project status X-UID: 93 There's less to see now as work is primarily systems integration. I'm trying to have confidence in the electronics and computer boards before they go back inside the aluminum enclosure. Before that happens, I plan on soldering all of the discrete components in place. A large number of unexpected things have happened since last Friday. The hacked USB card does seem to work with the UHCI USB 1.1 driver. But for a little while it disappeared from the PCI bus. After I pulled the card out and let it sit for a while, it started working again. It's magic. Aside: I am very superstitious with technology. I apologize to machines and tools if I have mistreated them lest they take offense and deliberately fail in revenge. I don't think this is any stranger than Japanese engineers holding pagan rituals to honor components that have been tested to failure. Another oddity is that an ext2 filesystem on a flash card can appear empty (no files in it) while having very little free space. At some level, the IDE bridge inside the compact flash card exposes the alien nature of the device. It is not a hard drive with true random access. It is a solid state device trying to level wear and yet provide an electrical and logical interface that appears like a hard drive. So weird things happen. The solution is to start over and remake the filesystem. The USB webcams I chose appear to work o.k. I will use two. My plan is to dedicate one for watching two horizontal laser lines. It will determine range on the two planes defined by the lasers. The other webcam will be used for optical flow. A conventional vision approach is stereoscopic correlation. This is too computationally intensive to be done with the cameras and computers onboard the the robot. It also may be overrated in real world situations. I've noticed that most field robots today seem to rely on pulsed time of flight laser rangefinders. Specifically, one of the DARPA Grand Challenge teams that relied on a very high performance FPGA based stero vision system last year now is using a SICK laser rangefinder. That tells me that they were not able to get enough performance purely from stereo correlation so had to adopt a more conventional approach. Here is short status. Some of the 0% items aren't nearly as difficult as some others. But I'm starting to have more perspective about my real ability. It is just like my day job. I can hit crunch time and put out massively for short periods. But it doesn't seem to pull a project along at a greatly accelerated rate. It just keeps it on time. vehicle platform frame, suspension and steering 90% motor mount 100% battery mount 0% electronics mount 0% electronics motor control board 90% single board computers 100% usb card 100% batteries and fuses 0% enclosure 100% GPS receiver on robot 0% 900 MHz radio on robot 0% GPS receiver on laptop (requires RS232/5V TTL translation) 0% 900 MHz radio on laptop (requires RS232/5V TTL translation) 0% wifi router on robot 90% webcams on robot (2) 25% lasers on robot (2) 0% Sharp IR rangefinders on robot (2) 25% piezo gyro on robot 25% XM-214 5.56mm minigun module just kidding firmware/software microcontroller 10% linux compact flash distribution 90% device drivers 90% robot software 0% base station control 0% I think that I can only estimate schedules accurately when I am tired yet with good morale (like now). Realistically, this robot won't be rolling until next January or February (only four months away - not much time, must push harder). That fits in with a winter timeframe. It is cold here in Dallas then. That is good for electronics and also keeps people away. The people reading this who are in the Dallas metro area, I invite you to a field demonstration then. A location with minimal spectator and police issues is desirable. I'm going to be loaded down with 50 to 100 pounds of very bulky robot and equipment. I don't want to have to worry about logistics and security in addition to the robot just working. I can think of some park areas where I could operate near my car. But I'm also thinking that it might be best to obtain access to private property where these issues go away.