Little Green Machine III
XR-B4
| Processor |
ATMEGA16
on ARC1.1 PCB |
| Sensors |
2 GP2D12, 2GP2D12,
1 SRF-05 Sonar,
1 UVTRON,
1 Pyroelectric |
| Actuators |
2 DC Motors with 2.5" Wheels. |
| Power Source |
9.6V 1600mAh Battery |
| Frame |
1/4" PVC |
Xr-B4 is my entry to the 2008 Trinity Fire Fighting
Contest. He finished 16th out of 42 robots, completing only 1 of 3 runs. On the first run, all was well until
he entered a room with contact paper on the walls. The contact paper tricked the IR sensors and he exited the room
at a really bad angle, crashing and getting stuck while trying to turn. The second run ran perfectly. On the third
run, he missed a turn for reasons unknown.
B4 used a hybrid architecture, combining a high-level planner with low-level reactive behaviors. The planner used
an a priori topological map of the arena. The software was written in C and uses the AVR-GCC compiler.
Several problems were still plaguing the system on game-day. The IR sensors were fairly reliable, but I still had some
issues. I am currently looking into Bayesian methods to solve this problem, although it is possible that I will switch to
predominately sonar sensors (even though thier speed is limited to about 20 updates a second across all of the sonars on the
robot).
|
|
