Anthropomorphic Robotics Project/Brainstorming
This is a rough draft.
Efficiency is the key component. Efficiency of code. Efficiency of mechanical design. Of energy use.
For sight, breaking things down into chunks. Maybe, have vision translate things into 3D, robots then process and orient themselves in 3D space in bits, and this translates into mechanical movement. But would that take up too much computing power? Could algorithms be written efficiently enough?
Air muscles - could they be designed well enough to make them aesthetically pleasing, but also powerful? Noise seems like a problem with air muscles. Also, if they are not designed well, they may be jerky in movement. How far along is the design of chemically driven artificial muscles? The cost and the state of the art seems prohibitive.
Suppose you tell a robot to do a task. It must break that task down into chunks, and create an executable system of steps to complete the task. Language processing is a must.
The robot must look non threatening. The robot must look at least like nothing out of the ordinary. You see a chair, and it looks normal - just like a chair. You see an iPod and it looks nice, it looks expensive, but it is just an iPod. It is functional. A robot must be like this.
What ways could voice synthesis be made to sound okay? Not like a robot, but like a friendly, anthropomorphic-like voice? Why is it so robotic sounding at the moment? Maybe this is a good starting point because decent code may exist, and it may be a necessary first step that requires little funding.
- computer vision
Matching materials to objects. Instead of using texture, sheen or other properties of a surface to determine the material, discover what the object is that is being examined and then reference what objects of this or that sort is usually composed of....