ME218b 2010 : Team BAP
Hats Off!
Project Goal Summary:
The goal of this ME218B project was to travel about an area and knock off hats assigned to us by a target commander through SPI. A round of competition lasts two minutes, during which the robot will attempt to attack eight different targets and then the robot will return to its home corner. A robot got 1 point for every hat it knocked down that it was intended to attack. For every hat the bot was not meant to hit, it lost 2 points.
The goal of this ME218B project was to travel about an area and knock off hats assigned to us by a target commander through SPI. A round of competition lasts two minutes, during which the robot will attempt to attack eight different targets and then the robot will return to its home corner. A robot got 1 point for every hat it knocked down that it was intended to attack. For every hat the bot was not meant to hit, it lost 2 points.
Figure Showing the Competition Arena Layout and Hat Locations
Our approach was to build a robot that could dead reckon from one location to the next. This required handling several specific challenges. First, we needed a super reliable drive train with as little slop as possible. Next we needed to control the motors tightly with as little slipping as possible. Additionally, we found that actively correcting our trajectory was critical as we slightly veered off target. Finally, dead reckoning can not be considered a reliable solution (especially after a collision), so we used IR rangers to square to the walls at a distance and reset our pose when necessary. To knock the beacons over, we simply used nerf balls at the end of an arm actuated by a servo.
General Bot Overview:
Drive Train: The final version of our bot relies primarily on a timing belt drive train, rigid couplers, and a slot allowing the timing belt to be easily tensioned. Because our drive train uses rigid couplers, we experience zero backlash, except for the small backlash in the gear head of the motor. Furthermore, our drive system uses easily accessible bolts, allowing for quick assembly and disassembly.
Position Sensing: To monitor position, we intended to implement a plan that used a combination of a particle filter and dead reckoning. Unfortunately, due to time and processor limitations, we switched to using our IR rangers as a method of squaring off to the walls by finding correct minimums. This corrected one coordinate direction and the orientation of the robot. Interestingly enough, the robot was usually capable of attacking 8 targets and return home using only dead reckoning, mostly due to the minimal backlash present in the drive system.
Hat Hitting Method: In order to design a robust, reliable system, we opted to use arms. By doing this, we placed more emphasis on building a good drive train and global coordinate system and less emphasis on calibrating our hitting method. Because the arms were controlled by servo motors, it was incredibly simple to design and build our hitting system. This gave us a simple, effective, reliable system and it allowed us to focus on other subsystems of the robot.
Drive Train: The final version of our bot relies primarily on a timing belt drive train, rigid couplers, and a slot allowing the timing belt to be easily tensioned. Because our drive train uses rigid couplers, we experience zero backlash, except for the small backlash in the gear head of the motor. Furthermore, our drive system uses easily accessible bolts, allowing for quick assembly and disassembly.
Position Sensing: To monitor position, we intended to implement a plan that used a combination of a particle filter and dead reckoning. Unfortunately, due to time and processor limitations, we switched to using our IR rangers as a method of squaring off to the walls by finding correct minimums. This corrected one coordinate direction and the orientation of the robot. Interestingly enough, the robot was usually capable of attacking 8 targets and return home using only dead reckoning, mostly due to the minimal backlash present in the drive system.
Hat Hitting Method: In order to design a robust, reliable system, we opted to use arms. By doing this, we placed more emphasis on building a good drive train and global coordinate system and less emphasis on calibrating our hitting method. Because the arms were controlled by servo motors, it was incredibly simple to design and build our hitting system. This gave us a simple, effective, reliable system and it allowed us to focus on other subsystems of the robot.
