AEPi at Carnegie Mellon marches to the beat of a different drummer. You know, that drummer in marching band, who is really nice and quiet but kind of smells weird. Yeah. That guy.
The boys of AEPi decided that Friday night would be a perfect opportunity to whip out the power tools and begin exploring the depths of RoboBuggy.
Mike Zankel and Brandon Sherman took lead of this endeavor.
 |
| Mike Zankel is in the stripes, Brando is in the light blue, and Nate is hunched over |
Nate and Alex watched in horror as Brando took an angle grinder to the rusted bolts in order to free the brains of the RoboBuggy.
After 3 hours of work, when the sparks and debris settled, the ancient brains of the RoboBuggy were unearthed.
Here is the full Inventory List:
1 – 150MB Flash Hard Drive and Ribbon
1 – Magnetic Compass
1 – Unkown home made relay component
1- Intel motherboard
1 – Pentium III processor
1 – ethernet card
1 – graphics card
1 – video decoder
2 – Expansion cards labeled I and II
1 – Shockmounted board
3 – sticks of RAM
1 – 12V 18A*Hr Lead Acid Battery
1 – 12V Computer Power Supply
1 – VGA Adapter
1 – 25 pin to ethernet and 4 pin adapter
1 – Optical Encoder
1 – Enormous Seiko Servo for the front wheel
1 – Pneumatic Braking System
1 – 6mm Color Camera
1 – Video Transmitter
1 – Logitech Wireless Mouse Adapter
2 – 37 pin input boards labeled CTR #1 and 2
1 – Cigarette Lighter Adapter and Power Panel
1 – 8 Channel RF Reciever
3 – CIA Buggy Wheels
1 – Steel Push Bar- Assorted RF, Ethernet and Power Wires
Here is a picture of the home made relay component we found. If any of our loyal readers have any ideas what this is, please comment.
 |
| ??? |
The final result:
 |
| RoboBuggy ain't go no smarts no more |
The Next Steps:
1. The two components that we are focusing on right now are the laser gyroscope and the optical encoder. Nate is currently playing with them in order to figure out their basic function.
2. We're also trying to figure out how to mount a camera to our normal buggies in order to get data during rolls. We're trying to find a camera that is cheap but would provide usable data, so we're looking for one that will be very close to the cameras we'll be using on the real deal.
Although we're planning on rolling RoboBuggy in exhibitions during this year's races, we want to make this buggy adhere to all the rules of a normal buggy. Our goal is to make RoboBuggy a competitve buggy in the near future.
Here is the complete rulebook that I found online from 2009:
One idea to increase the buggy's vision signals is to provide extra flaggers around the course that will help tell RoboBuggy where it is. This is within the rules (see the last paragraph of section 8.1.3).
A downside to adhering to the rules is that we would like to put a GPS unit in the buggy. This is a problem because section 10.5.2 under "Buggies" state that the driver can't have any communication to the outside world that isn't already accessible to them, and it specifically states that telemetry units are prohibited. Now although a lot of our other components may provide a little bit of haze within the rules, we can make the argument that we're simply mimicking all the senses of the driver, IE sight, speed and direction. Unless Sweepstakes says otherwise, we will need to figure out what we want to do with the GPS, as it can be incredibly helpful.
In addition, if you would like to learn more about RoboBuggy, simply see the project, or learn about regular buggies, feel free to either email me at alexklarfeld@gmail.com or simply stop by AEPi! We don't mind sharing our buggy "secrets" with interested parties.
-Alex
Is an autonomous buggy your only goal? It seems like having a driver operating the buggy with remote control seems like it could satisfy those rules. I didn't read over everything, but from the passages you stated, that would not break any of those rules if you consider the driver to be the person steering--not the buggy. Granted, an autonomous buggy is cooler...
ReplyDeleteIf the "magnetic compass" was a black box mounted on top of the battery, it is actually a 1-axis laser ring gyro.
ReplyDeleteThe homemade relay board controls the brake and servo. It operates 2 solenoid valves, 1 to let air from the tank into the brake cylinder, and 1 to let air from the cylinder out to atmosphere, giving fine control over braking force, but with some lag. It had a serial? interface with the computer and an RC servo interface with the radio receiver, and an output to to the steering servo. Normally, you'd get the sum of whatever braking the computer & remote operator wanted. It the computer crashed or the RF receiver lost the signal from the remote control, it would fully apply the brakes. The pneumatics were arranged so that if the battery failed, the brakes would automatically apply - the only danger was forgetting to fill up the air tank.
I think the mystery expansion cards are digital & analog I/O boards. These were used for reading the signals from the steering angle encoder, the gyro, and the radio receiver. I forget if these were routed through the relay board or not. (the design changed a couple of times)
Despite the steering servo being massive, we managed to destroy one by accidentally having it alternate from full-left to full-right as fast as it could - the inertia of the steering & wheel overloaded the gearing in the servo. The steering had the stupid design feature that the wheel could be steered far enough to run into parts of the frame. The relay board would normally limit the angles to the safe range, but it didn't always work when driving manually.
The computer hardware was state-of-the art for 1998, pathetic as it seems now. The reason behind the massive lead-acid battery is that lithium & NiMH batteries weren't commonly available then and lead gave much more warning that it was running low than NiCd.
We never seriously thought about GPS back then, because affordable receivers were neither accurate enough, nor provided rapid enough updates to drive by. And getting a $12K DGPS unit would have made the task trivial.
ReplyDeleteIn the end, vision worked very well (better than a human driver) for line following, but not well at all for detecting landmarks. The rapid change from shade to shadow & back made things very difficult for the vision system because the image would go all white or black for a few frames. Plus, bumps would jar the mechanical iris in the camera, causing the same effect.