One huge project I'd like to build is a full size humanoid biped or "android". There are numerous projects like this around the world, complete with incredible costs in building them. I do think that with current technology, it is just possible for a hobbyist to attempt and perhaps complete something in a similar vein. And several are trying, working at their idea of a biped, perhaps as a small servo driven model. Though such a model serves as a type of proof of concept, I'd really prefer to start with something "in our own image". Which of course brings the price back up. So the answer seems to be to start small, with a section of an android, perhaps the head, the legs, an arm, or even just a hand.
The above hand picture isn't a development of mine, but it does show some of the detail and complexity needed in something as simple as a hand. Which further points to the divide and conquer prospect. The whole of the project does need to be considered though, at least to get some idea of the end android. As such, what I'd like to accomplish is :
- bipedal walking
- obstacle avoidance and recovery
- general location
- self recharging
- object manipulation
- verbal communication
- aesthetics
Each section may seem strange, but each represents a milestone in one way or another. Bipedal walking on it's own is a difficult task, as shown by the years and dollars spent on the various projects around the world. And just because something can walk doesn't mean it won't run into things, so there needs to be some type of local sensory input combined with recovery. General location is helpful in ways we know where we are. Right now, I can tell you that I'm in front of my computer typing this. I don't know my precise xyz coordinates nor my orientation in space, but it's in my work area. I know how to get to the bathroom from here and this is what an android should be capable of. This ties into self recharging, specifically being able to find an outlet (general) and plug in (local, from obstacle avoidance). Once objects like outlets can be found or recognized, then they can be retrieved or manipulated. Communcation includes being able to accept simple commands on up to AI implementations for conversation. And finally aesthetics is just simply a given. We all like things that are beautiful, in one form or another, not necessarily a mass of wires (though that can be an appealing art for some), but an outer form that completes an android. Figures of androids have been shown in many forms, from C3PO in Star Wars to Data of Star Trek. In the end, it must have general appeal. And hopefully, eventually, it will.
|
The DPRG Roborama contests provide a test of ideas in the form of friendly competition. Most successful robots use some form of odometry to travel a given distance in a contest and return. I'm hoping to build a successful robot that does not directly use odometry, but can simply wander through the contest area with just a "good idea" of where it is at. And hopefully be able to collect cans for the Can Can portion of Roborama. All controlled by a Basic Stamp 2. And then I wake up. It is a possibility for a robot to wander through the contest area visiting both check points and returning to start. Continuing to wander until all 6 cans are brought back could be implemented. Combining these goals is the key. Of course, the motto of the DPRG is,"it's harder than it looks".