Adaptive LEGO robots. A robot=human view on robotics

摘要

In many applications, robots are viewed as a machine. This has resulted in interaction and actuation which is characteristic for machines. When constructing adaptive LEGO robots, we take another view, namely that the robot should resemble a human (or a biological creature) rather than a machine. This has implications on the interaction, actuation and control of the robot. I describe how the robot-as-human approach is investigated in a number of LEGO Mindstorms robot applications. These include making facial expressions, which allows a LEGO robot to express internal "moods", and thereby we might achieve a better human-robot interaction. Another application is the Adaptive LEGO Pet Robot. The Adaptive LEGO Pet Robot's control is based on a modular behaviour system, where a number of the modules are evolved neural networks. Further, the Adaptive LEGO Pet Robot has a number of internal drives such as restlessness and hunger, which allow the robot to react on the internal drives. The human-robot interaction is facilitated by allowing the human to train the LEGO pet robot (rather than to program the robot) to make associations between spoken words (via speech recognition) and evolved behaviour. The Adaptive LEGO Pet Robot is an example of scaling up evolutionary robotics to complex behaviour by combining evolutionary robotics with behaviour-based robotics.