Method of research

Robotic agents differ from purely software agents, since the environment in which they are expected to operate is not just an abstraction or a virtual world, but the real world. However, the real world is far more complex for us to design a simple and general model. When we develop integration technology of robotic system and each fundamental technology, we must evaluate new functions and their limitations through real world application. When putting in practice concrete issues, problems may seem countless, but then new solutions to these problems may also come out. Therefore, it is important to repeat experiments in a real environment as much as possible for developing a robotic system. This "experiments in the real world" is our first leading methodology, and we call this approach "Robotics as an experimental science" or "Experimental Robotics". Of course, each fundamental technology is important for useful robots working in real environments. However, it may be not possible to really know what kind of elements might be necessary only by an approach starting from fundamental technology. So, we are trying to bring out the following approach. At first, we decide the practical task the robot should realize and the environment where the robot should work. Then, we develop a real robot to realize such task, to discover the required fundamental technology through the developing process and researching on them. This is the second leading methodology of our research. We call it "Task oriented approach". To perform our research based on these methodologies, we need robots as our experimental tools (robotic platform). As a research tool, the robotic platform is expected to be a commonly shared tool with generality. That is the reason why we are also developing and building our own mobile robot platform, and we have called it "YAMABICO". As our task-oriented research methodology may be associated with a top-down approach, on the opposite, our aim at developing a robotic platform may be related to a bottom-up approach. The original characteristics of our research are our experimental approach to realize our final goal : "Automatic machines well-working in the real world" combining these two different methodologies.