Demonstrated
at Shikoku Research Center of Japan National Agricultural Research Center for Western Region.
Intelligence can only be determined by the dynamic interaction with the world. Recent studies
in artificial intelligence and robotics suggest that even very simple biological creatures exhibit
desirable functionality and boast far greater robustness than human-designed systems. This project
focuses on using the biologically-inspired approaches to design an evolutionary autonomous mobile
robot that is able to navigate in a poorly defined environment. The implementation of complex
artificial intelligence systems can be approached by decomposing the global tasks into simpler,
well-specified behaviors which are easier to design and can be tuned independently of each other.
The behavior-based approach is chosen to build the control system of the robot. Robot behaviors
can be implemented as a set of fuzzy rules which mimic expert knowledge in specific tasks.
Behaviors usually emerge from implicit knowledge of the underlying process that can be converted
into a set of linguistic variables and fuzzy rules. A vision-based landmark recognition system
for robot navigation tasks is implemented as the highest layer. A novel search method, based on
genetic algorithms for pattern recognition in digital images, is proposed and implemented in the
developed mobile robot to generate desired behaviors.