One year ago I posted an article about my new autonomous boat. Well it is now almost complete. The construction of the hulls is now mostly complete and I have finally got the new ARM based single board computer working. Over the holidays I plan to complete the prototype and have it in the water.

Hull Design and Construction

Okay, so what is new in this prototype? My last prototype had two rather serious flaws. First was the hull design; it was essentially a barge. This meant gross directional instability. In waves it would be easily pushed around. It was also constructed from balsa wood which was very weak and resulted in persistent leaks. Not so good for a boat.

I decided that for the new prototype I would build a catamaran. This would both aid directional stability and allow a larger surface area for solar panels. I also decided to construct the prototype from something a little more robust; plywood.

After doing some research online I found a design for a simple catamaran that was free, along with detailed construction advice. The construction method was called stitch and glue. This approach was simple to execute and didn't involve exacting woodwork; exactly the kind of construction method I needed.

The primary construction materials in this approach are plywood, epoxy and fibre-glass. First the plywood panels are cut out. Small holes are drilled along the edge of each panel, and wire or plastic ties are used to join the panels into the shape required.

Once the ties are all in place epoxy and filler is used to in each join, and then fibre-glass tape soaked in epoxy is placed over the top. All the panels are also painted in epoxy to seal them from moisture. Once the epoxy sets the ties can be cut away.

To finish the exterior I have used an oil based undercoat and orange enamel paint to improve visibility on the water. The end result are hulls that are strong and watertight.

Electronics Package

The hull is however far from the only change. The previous prototype used a OpenMoko phone as the central computer. The advantage was that the OpenMoko had communications and GPS built in. However it was not without issues. The GPS required a binary driver, and it was increasingly difficult to install software outside of the standard install such as python serial.

There was also the issue of power, and how that if the OpenMoko powered down for whatever reason it would not automatically reboot once power is restored. That was a pretty serious problem for a system that will be out of reach.

As a result I did some research for a single board computer that required low power. It had to be able to restart when power is reapplied, and ideally run efficiently on a 12 volt supply. It also had to run a standard Debian distro so I could install anything from the standard repository. I found the TS-7260 from Technologic Systems.

The TS-7260 has some very desirable features. It has a switch mode power supply that can take anything from 6 volts to 20 volt. Because it's a switch mode supply it doesn't lose 50% of the power in heat like a standard voltage regulator.

It also has built in software switches to control power to USB. This means that I can switch off power draining USB devices at night. I'm still thinking about shutting everything down and using a simple timer to reboot after a period, but this may not be required if we can reduce the power consumption to a reasonable level over night.

Getting the TS-7260 up and running was a bit of a mission, with issues around getting many basic systems going. After a few months of trying to get the old python application running I finally got it running in late December.

In the last prototype I had only one USB device, the PhidgetServo. This has increased somewhat in my new prototype. The GPS is now an external device rather than built in. The communications is now done through a separate USB 3G device. I have also added a PhidgetInterface to handle analog sensors and a USB compass to determine orientation. In my last prototype I had real problems with trying to establish where exactly the boat was heading. So we have gone from one USB device to five.

The power is now monitored through two current sensors showing how much power is collected and used. I have also written a python module to interface with the . This was based on a manual for the compass that was not officially released. It has a rather interesting binary protocol.