The Metal Robot: What Went Wrong

Rough Schematic of the Metal Robot

A number of factors contributed to the failure of our metal robot platform, most of which can be reduced to poor planning and lack of precise design and engineering. The platform began as a vague vision in John's head; key design decisions early on were based on estimates and assumptions regarding the sonar avoider robot Corey built. The weight and size of the robot were not precisely planned; rather it was assumed that our platform would not weigh much more than three times that of Corey's robot. This led to the decision that the 1926Y motors would be more than adequate. These motors were the most powerful available in the DSL, and it was assumed that of course they would do. Due to an error in documentation, we were also led to believe that these motors were 100rpm, when in reality they were only 44rpm.

The first hint of problems with the metal robot design came during the machining of the metal parts. John had previously worked as a machinist, and had made an estimate of the time required to machine the parts based on this prior experience. Unfortunately, this estimate was flawed since it assumed ample time each day to work on the machining, as well as easy access to the mills and tools required. John soon discovered that it took considerable time to collect all the tools he needed as well as obtain access to the school's machine shop. The machine shop also had restricted hours, which prevented John from working as quickly as he had planned.

It soon became apparent that with the slow pace of machining, it would be difficult to complete all the parts necessary for the planned four-wheel steering. In addition, the gears purchased proved to be too small; no careful measurements of what size gears would be needed were made prior to purchase and order. Cheap press-fit bearings turned out to be cheap for a reason and would have to be replaced if steering would work. This meant more money to order new gears and bearings, as well as more time for delivery and machining.

We sat down and discussed the problems we faced with steering, and at this time we made our second fateful decision. In order to save time and money, we would build the robot to turn by inverting the direction of each pair of motors on each side, or 'tank-steering'. This would not only save us the time, money and effort of implementing real steering, it would also make the interface with the jStamp much easier because it would cut down ont he number of H-Bridges needed from 12 to only 4. It would also make the software class for controlling the motors easy to implement and use because we would control the robot's steering with DC motors instead of the stepper motors we had for steering. Again, we made no design measurements to ensure that this decision would lead to success; it was assumed that the DC drive motors would have plenty of power because they were the most powerful that DSL had.

Machining proceeded and as the robot was nearly ready, we began working on the interface to the jStamp. Test circuits were built and signals sent from test classes to the H-Bridges. Bruce gave us a number of good suggestions for getting the number of signals we needed from the jStamp and we began planning a motor interface class to use in the main robot control class. The large number of wires needed to connect the H-Bridges made soldering difficult work, but it was still doable.

During this same period, the robot began to take shape. Once it was sufficiently assembled, we wired up the motors and batteries to test the robot's speed and steering performance. We discovered that not only was the robot much slower than expected, but it was also totally unable to turn using 12V, as well as at 24V. Our idea of the robot's speed had been based on the assumption that the motors we were using were 100rpm, but we discovered that the DSL documentation was incorrect and that the motors we were using were only 44rpm. We assumed the inability to turn was due to inadequate motors; Bruce suggested that the problem might lie in loss of power in our electrical connections and that we take measurements of the actual performance of our motors to see where the problems lie.
NEXT PREVIOUS