In 2004 a team of 5 of us decided to build an omni-directional electric vehicle called the Gravity Bowl. Affectionally known as the Dog Dish, this car was built to take to the Burning Man arts festival in Nevada.
Thus started a nine month project of extreme labor and complexity which took over all our lives, and particularly mine, as I had volunteered to design and build all the electronics and all the control software to drive the vehicle. I also did some of the chassis welding. Kevin MacDonald was the mechanical designer. The rest of the team pitched in handsomely, and they were Geoff Silverton, Brady Forrest and Lars Liden.
Here you can watch a short documentary showing the process of building and running this car at Burning Man. At the bottom of this page you can see a gallery of photographs of the car during construction and running at Burning Man.
The initial concept for this electric vehicle was for a circular couch where everyone faces inwards encouraging a social experience. It was intended to be omni-directional so that it could float like a flying saucer in any direction. We wanted to build it so that all the wheels could steer so that it could both rotate and translate at the same time.
We decided to build it in six sections that could be disassembled when transporting it. It was to have three driven motors and in the end ran off of 3 banks of 4 marine batteries. We applied for an obtained an arts grant from the Burning Man arts foundation, which took off a little of the financial load.
Initial design concepts had a wheel in each section with 3 idler wheels, but this turned out not to be necessary, and the additional segments were used to hold the batteries and power control electronics. The car was 12 feet in diameter and designed to carry around 8 adults. However there were issues in practice with the motor drives – particularly with the gear ratio – which limited its ability to get out of grooves in the ground when a large number of passengers were involved.
The full construction was built out of steel welded together and then covered with aluminum panels that were nicely burnished. There was a rather nice top railing that was painted black and had a sparkly finish. The finish of the machine was rather good and there were comfortable and tasteful seat cushions. This made it great as a bar table.
In addition there were a handsome set of lights, both illuminating the ground and offering down lighting of the sides of the panels. There were also white LEDs that sequenced around the bottom of the outer rim so that it looked pretty great at night.
The steel construction and large number of batteries made the vehicle rather heavy and this contributed to the problems that we experienced at Burning Man, which involved the breaking of chains in the wheel units during moments of significant stress. In addition we had issues of dust ingress into contacts on the buttons that were used to drive the machine and these failed because they were not adequately sealed. The amount of work involved was very great and the car was not finished until half way through the festival, and I spent a significant time writing software to get it up and running at the event. However, it was a great project and we did get some fun out of it.
The control system for this car was overly complex and I insisted on designing it all in detail from the ground up. Retrospectively this was a mistake, but I had to learn this lesson for myself. Much time would have been gained if I had purchased off the shelf motor control units instead of trying to design them on my own. In the end they worked perfectly, and I learned a lot about power electronics for driving 200A drive motors, but it was a highly stressful endeavor.
Before building the car I did an analysis of the motion modes that it could go through, and how it should be controlled – essentially it was a three wheeled vehicle where each wheel was free to steer independently, so controlling its behavior was going to be a little complex. You can read my writeup here. In addition, I did a writeup of the control system for controlling the steering and drive motors which is here.
In overview, the motion control system consisted of four micro controller boards all based on the DSP56F805 dual motor controller. This was a 16-bit 80MHz DSP chip and was sufficiently powerful for this application. Three of the boards were slaves and associated with the wheels, with each board controlling one steering motor and one drive motor, and the fourth controller was the master used to coordinate the motions and interface with the user. This controller was also tasked with sequencing the lights. They all communicated using the CAN network which is a little bit like ethernet.
Each of the slave controllers talked to a power electronics board with two 200A 48V H-bridges to control steering and driving. These had current/force feedback using hall effect current sensors and had quadrature encoders used to sense the wheel rotation and steering angles. They involved two or three nested PID controller loops for current control, speed control, and in the case of the steering motor, the geared-down steering angle.
All the boards were four layer PCBs designed using EAGLE and were built entirely using surface mount components which were manually soldered. This was a bit of an arduous task, but turned out ok.
I learned a lot about building high power motor controllers. In order to validate the design, I had to do a full SPICE simulation. I found it necessary to use some fairly expensive capacitors to handle the local ripple load – something which they often skimp on in commercial units.
The drive motors were these pancake-construction E-Tek motors. I found that they had very low inductance, which turns out to be bad for motor control because it means that there is a very high ripple current during operation and it is much harder to control the motor current. To help with this I made some high current toroidal series inductors on ferrite cores which worked quite well to smooth things out.
The whole electronics system was rather over-designed, but worked quite well. Its just that it took a very long time to build and write all the software from the ground up and was highly stressful when more and more problems came up during the time leading up to the event. Much appreciation is due to my boss at Microsoft at the time, Matt Uyttendaele, who gave me a pretty large amount of freedom to complete this project while I was supposed to be working full time.
The main controller used the same circuit as the slave controllers, but had slightly different components soldered onto the board. You can look at the schematic here. It had a port for an LCD display and buttons used to control and drive the vehicle. It was intended that the art car should be driven by a game controller handset but this was never completed in time.
Instead we had to rely on buttons on the main control box which had actually been added as an afterthought. These were ok for some time, but became unreliable after a few days on the playa in dust-storm conditions, making driving impossible.
After we brought the vehicle back from the playa, it was used for a number of years as a rather nice cocktail lounge couch in a local speakeasy, but eventually was scrapped due to storage costs.