The robot's evolution over time

Sun, 27 Feb 2005 03:41:28 -0600 (CST)

I remember thinking that the robot could be quickly put together by taking shortcuts to avoid the mechanical problems inherent in a suspension and drivetrain. For the suspension, fiberglass rods epoxied into metal pieces supported the wheels. As it turned out, these rods were far too stiff. For the drivetrain, a flexible shaft originally intended for driving screws with a drill both took lateral loads off the drill bearings and allowed the drill motors to be offset. The drill motors themselves were held together with zip ties. I was never very happy with this design and fortunately epoxied one of the metal pieces one inch off-center (accidentally) which finally killed it off.

Mon, 28 Mar 2005 22:53:20 -0600 (CST)

For about a month, I explored the design space with TurboCAD. I had never used CAD for anything in 3D before so was constrained largely by what I could draw. Everything was bolted together, leading a coworker to comment that my designs were “Victorian”. I pretty quickly settled on a hardtail configuration due to the complexity of the drivetrain. There was only a front suspension. For this, CAD was invaluable to visualize the parts and get them to clear with travel. The vehicle also kept shrinking during this time. A requirement was that it fit inside the trunk of my car. I remember wanting it to be very “dense”.

Mon, 30 May 2005 23:39:15 -0500 (CDT)

By this point, the mechanical design was largely set. The front suspension worked pretty well. The idea of using a spring in tension to pull the swing arms down seemed to work. A conventional suspension uses springs in compression to push the swing arms down. But I didn't want to deal with a shock tower as on a dune buggy. The front wheels used two steel tubes epoxied together to mate with the kingpins. One of the epoxy bonds gave way when I picked the robot up by its front wheels. This sort of highlighted how not having any welding capability was a handicap. There are many design elements and fabrication approaches that don't work without welded metal.

Sun, 5 Jun 2005 18:20:35 -0500 (CDT)

The new kingpin blocks were made of single pieces of aircraft aluminum instead of two pieces of steel epoxied together. I took it out into the parking lot and rolled it around to get a feel for it. Fortunately, it seemed to track straight. There wasn't any shuddering or wobble I could feel. I did note that the connection between the rear wheels and the driveshafts could be problematic. Under enough torque, the wheels would naturally tend to break free. The other thing I noted was that people were staring at the robot even in its incomplete state. Cars would stop in the middle of the road. This told me that field testing requires planning to avoid crowd control issues.

Fri, 17 Jun 2005 02:18:04 -0500 (CDT)

The steering mechanism had been added. It used one of the flexible shafts again. Later, this would be improved and the flexible shaft found to be unnecessary yet again. For mounting electronics and batteries, a shelf made of steel tubes and 1/4” aluminum was made. This turned out to be too heavy. It was also pretty cramped between the drill motors and wheels to fit very much. But weight was the main issue. The shelf was overbuilt and could itself support several hundred pounds – which would crush the robot.

July 2005 to November 2005

Many months transpired while working on the electronics. I eventually broke down and purchased an oscilloscope which saved me. The first motor control board with three H-bridges on it was discarded and salvaged for parts. The second motor control board was much simpler and actually cheaper.

Wed, 4 Jan 2006 22:35:22 -0600 (CST)

By this point, all of the electronics had been function tested at least. The two unknown areas, whether the motor control board could handle higher power levels and whether the webcams were grabbing good images, didn't worry me. But I was worried about mounting everything on the vehicle. I couldn't visualize how everything would fit. Another mockup was necessary. This indicated the two problematic areas: battery packs; lasers and webcams.

Sat, 4 Feb 2006 14:54:48 -0600 (CST)

The robot hardware is completed.

The two problematic areas identified last month did turn out to cause trouble. The batteries took more work than expected. And mounting webcams, distance sensors, and gyro went through several iterations before reaching the final configuration.

There are still some problematic areas. The robot is heavier than anticipated. The swing arms in the front suspension should ideally slope up from the wheels, not down or level. The locking nuts connecting the rear wheels to the drive shafts come loose easily. And the robot is difficult to transport due to the size and weight.

Sat, 11 Mar 2006 23:06:12 -0600 (CST)

A network camera is added to the sensor bar for a total of three optical cameras in the “Berkeley configuration” - one fast monocular camera for road following and two cameras for stereo obstacle detection. Minor modifications to the drivetrain and steering too. The basic mechanicals have been checked out as demonstrated by this remote driving test: mvi2826.mpg