The field of Evolutionary Robotics (ER) starts with a fixed robot platform which has a computer brain connected to sensors and effectors. Different control programs or ``brains'' are tested by downloading them into the robot and evaluating its performance, either in the real robot or a simulated version [37,39,22,67,72].
Among the difficulties ER faces are,
Evolution is limited by the pace and physicality of the robot. Making copies of a hardware robot is costly because they are not mass-produced. Also, the pace of time cannot be sped up.
The robot itself is designed by human engineers who engage in a costly process of designing, building, testing and repairing the robot. Commercial robots are available for research on a limited basis.
With a robot ``body'' whose morphology can not change, ER is limited to evolving control programs for the fixed platform. This represents a strong limitation, as we argue below, when compared to biological evolution where all behaviors are supported by morphology.
Evolution is limited by the pace and physicality of the robot. Making copies of a hardware robot is costly because they are not mass-produced. Also, the pace of time cannot be sped up.
The robot itself is designed by human engineers who engage in a costly process of designing, building, testing and repairing the robot. Commercial robots are available for research on a limited basis.
With a robot ``body'' whose morphology can not change, ER is limited to evolving control programs for the fixed platform. This represents a strong limitation, as we argue below, when compared to biological evolution where all behaviors are supported by morphology.