B. 2013 Rovers and Lessons Learned
The starting point for the 2014 SRR is the work done for 2013. The picture shows myself with the rovers and the home beacon in the robot pits during the 2013 competition. The robot and home beacon probably will not change much for the 2014 competition.
Early in May 2013, each competitor was required to send to the challenge a final written proposal and a video showing the robot performing certain actions. In the video and proposal you can see the way samples are lifted and carried and how the rover moves. The arrangement of the camera and components on top of the computer changed a bit between the video and the challenge. You also see how samples are taken off the lifter by brushes on the starting platform. There are three brushes that are 1.4 meters long. That would allow 3 samples behind each brush and one left on the last rover, assuming all ten samples were collected. There are additional videos including ones from the camera on the rovers during their SRR run.
The home beacon was a roughly 2 m x 2 m white panel edged with 8" of red duct tape, front and back. Fiducials on the front and back indicate to the rovers the orientation of the beacon and, thus, whether they were approaching the front or the back of the starting platform. At the top of the home beacon (not visible in photo) is a wireless router configured to provide WiFi for communication among the rovers.
The 2013 Mystic robot consisted of 3 rovers and a set of brushes that resided on the starting platform. Each rover is as identical as hand construction allowed. The base chassis, see photo, is a Dagu Wild Thumper 6WD with 75:1 gear motors. The motors are controlled by Pololu High-Power Simple Motor Controllers.
The control computer is from Mini-Box. It contains an Intel motherboard with dual Atom processors, a solid state disk, and a WiFi controller. The motor and servo controllers are mounted on the top of the computer chassis using the ventilation holes.
The servo controller is a Pololu Mini Maestro 18-Channel USB Servo Controller. The Maestro not only controls the servos but also reads inputs from a receiver for a key-fob transmitter, used to trigger specific routines for testing, and flashes the amber LED safety light. Vision input is a Microsoft LifeCam Studio web camera. The controllers and webcam connect to the computer via USB.
There are three servos used by the rovers. Two are in a SPT-100 pan / tilt from ServoCity that allows 360 degree vision around the rover. On the front of the rover is a ServoCity DDT-500 tilt mechanism. This tilt is used to position a scoop under samples and then lift them for carrying.
Each rover is powered by two 7.4v 10 mAh LiPo batteries wired in parallel. They ride in baskets between the wheels of the rover.
The major issue with the rovers for 2013 was software. It simply was not sufficient for the challenge. The reason, as is usually the case with an embedded system, was the hardware, specifically the platform. Early in May 2013 I had completed assembly of two of the rovers in their final configuration. During testing two problems appeared. When making tight turns on grass (1) the rovers would stall and (2) a voltage drop occurred which killed the computer. It took a good deal of time to find that the voltage drop occurred in an electronic switch used to power on the rovers. (One of the rules is that the robot is turned on by one switch which means the three rovers need some form of remote power control.) The switch was replaced by relays wired to latch as long as power was present.
The chassis I was using had 34:1 gear motors. These were selected to provide the 2 m/sec (or slightly higher) speed allowed by the competition. They did not have sufficient torque to turn the rover in its final, fully weighted configuration. New 75:1 chassis were obtained and software development time was lost in replacing the chassis.
Once at the competition the rovers were tested on The Quad at WPI. A problem appeared with the 'lifter'. As you can see in the video it tilts down, the rover moves forward so the lifter is under the sample, and the lifter tilts back to hold the sample in the carrying position. (A minor, easily fixed problem occurs when traveling on rough terrain - the sample slides of to the side and can drop off the lifter.) At WPI, the lifter kept driving into the ground, stopping the rover, bending the lifter, and doing diabolical things to the servo used to tilt the lifter. (The lifter is a wire tie rack.) I am not sure why it worked in Houston but not at WPI. One possibility is when fully loaded with equipment and batteries the rover is lower to the ground causing the lifter to dig in. The other is the difference in grass between the two locations. Seriously!!! Grass in Houston is Saint Augustine (typically) which is very dense and tough. The rover often is riding somewhat on top of the grass at the house. The grass at WPI is not as tough or dense (wimpy grass!) so the rover rides closer to the ground. I spent time trying to get the lifter to work at WPI without success.
The lifter needs improvement or I need another way of collecting samples. A problem is I cannot add much more weight or mechanisms to the rovers.
The Institute Park, where the competition is held, has a hill on the east and west sides. (These directions are based on the competition reference grid. In the competition, Salisbury Lake is south of the park with Salisbury St running east and west and Humboldt Ave going south.) Mystic Two ran along the Salisbury Lake and then turned to go up the hill. A little ways up the hill there is a 6 inch drop off and Two rolled over.
The rovers need an accelerometer to indicate when a tilt is occurring that could cause a rollover.
The major, positive lesson learned is that the rovers can easily handle the terrain of the park. Mystic Two ran from the center of the park, along the water line and only stopped trying to go up the hill. The other rovers did not travel that far but they were not stopped by the general terrain. (Mystic One just stopped for an unknown reason. Three ran into a rock.) The detection of the orange boundary also worked well. This contributed to Two's long run along the fence at the edge of the lake.
My conclusion is that the rovers are competitive in the challenge with the only open mechanical issue how to collect the samples.