Autonomous Astromech

From KallestadWiki

I want to build my own R2 unit. There are a lot of people building RC astromechs out there, but I haven't read about any autonomous droids. I think it would be a fun project. I'm sure it will take years.

A lot of the technology we need is already available in some form or another.

Here's a list of components that I think will be required:

• Linux based controller - easy to develop in perl/python/etc, easier to hack with lower cost than a PC environment.
• Voice recognition - not sure what linux solutions are out there. Google tells me:
  • Actually this page has a lot of information about voice recognition.
• Digital compass - usb enabled so I know what direction I'm facing
• usb Range Finders - for wall detection, learning about your environment, avoiding stairs. I found this I'm sure there are less expensive solutions out there to find. These one's look promising: robotshop
• video camera - for recording the environment, visual recognition for sign based commands, object recognition
• video projection - 3d holography is still not very prevalent, but I happen to have a vga projector that I haven't used in years.
• Verizon or some other cellular internet. I have an extra verizon card, but maybe a hack is necessary for good antenna location
• sound output - I like R2's voice, but there are some good human voices out there that are worthy for at least early training
• Robot arms - there are a lot of robot arms available out there. I'm thinking it might be cool if R2 could plug into a wall for charging and also talk X10 and other powerline based protocols.
• stun gun - for offensive ploys like fending off yoda
• power generation - solar energy, finding and pluging itself into a wall socket, maybe an alternator or some sort of internal power generator that captures movement based energy for restoring a percentage of expended power to extend battery life.
• GPS - for outdoor location referencing

Servos are a possibility for a lot of the moving parts, but I'm wondering if a gear based system may be a little more appropriate. A lot of servos makes for a complex electrical system. Maybe a mechanical system is more of a realistic solution. You could store energy in a coil or a spring for little movements and run a small engine using geared spindles like a 10 speed bike. Of course there's the issue of reversing direction. I wonder if an alternate set of gears could be used for reverse rather than finding an engine that is reversible. Shock absorbers could make movements more smooth.

I had a good idea for navigating single steps and curbs. A small set of aluminum rails could be stored in the feet. They could get pushed out automatically and dropped. R2 rolls down his makeshift ramp and uses an electromagnet to pick up the rails. A guidance system then feeds them back into the storage area. Another possibility would be to use body tilt and his third leg, but I have trouble imagining how that might actually play out.

Movement

I've been thinking about propulsion and it's a subject that is eating away at me. Do I opt for fixed position wheels, or rotating wheels? Independently driven wheels, with multiple engines, or a common engine driving the wheels using engaged/disengaged gearing?

What about stopping? Engine braking or friction braking at the wheels? A worm gear can be pretty effective at stopping, but the sudden stops could be disastrous at high speeds. How fast do i want R2 to go? And a bigger concern - is how accurate can I expect his movement to be? Ideally, R2 would map the world around him and once he learned an area he would be able to move more freely and faster within that region. The map is only as accurate as his sensors though, and the accuracy of his location given the data received by his propulsion mechanism.

Free spinning wheels for coasting? or constantly engaged ones.

These are a lot of questions to ask and answer. Part of the answer has to come from what I expect R2 to be able to do. The other part comes from budget constraints.