There is technological/finance gap between intelligent robots able to talk and singlepurpose hobbystyle creatures made of affordable robot building sets (Lego, MakeBlock, Tetrix, Arduino etc.). It’s still a science fiction scenario when some kid can build a robot, turn it on and start talking, assigning some realworld tasks which need some decision making. What I want to do is to implement this scenario and propose simple and elegant solution for addition of a talking and seeing brain to ”alive” any simple robot built from popular mechatronic components or bricks. To perform this task I developed a robotic brain (development name is ELFi) in form of Android Application.
Community of software developers is critical part of future of ELFi development . So I had to develop an extendable software architecture for simple injecting custom code in any layer of ELFi Brain. I decided to use Android Application Development patterns to solve this problem.
I had to develop a bridge between concurrent processes running inside Android OS and realtime processes runned on robotic bodies. Also I wanted to remove low-level and controllerspecific tasks outside of robotic brain.
I created a special intermediate software level named Algorithm Executor with perform such a separation. I configured and tested simple voice commands for simple tasks. Also I integrated external AI cloud for Natural Language Understanding. This gave ability to ELFi brain to recognize more flexible spoken constructions, recognize contexts and spoken scenarios. I developed prototypes to proof the ability to inject ELFi brain into different affordable robotic construction sets.
This software platform can be applied for various types of intelligent robots.