Hardware solutions for evolutionary robotics [chapter]

Dario Floreano, Francesco Mondada
1998 Lecture Notes in Computer Science  
Evolutionary robotics| as other adaptive methods, such a s reinforcement learning and learning classi er systems|can take considerable time and resources which require a careful evaluation of the hardware tools and methodologies employed. We outline a set of hardware solutions and working methodologies that can be used for successfully implementing and extending the evolutionary approach to complex environments, robots, and real-world applications. The issues discussed include the integration
more » ... simulation and real robots, design issues of evolvable robots, hardware requirements for incremental evolution, and hardware and software tools for monitoring and analysis. arti cial evolution. Capitalizing on these principles, we shall also discuss hardware solutions which can support incremental evolution, namely modularity and cross-platform compatibility. Finally, we shall address the issue of monitoring and analysis of evolved robotic controllers. Between Simulation and Real World Evolutionary robotics has its roots in simulations. The rst explorations in evolution of autonomous sensorimotor agents date back to the end of the Eighties, 3 when all studies were still carried out in computer simulations. A few years later, the appearance of more robust, exible, and user-friendly robots, and a general awareness of the limitations of simulation methods 2], created a strong motivation for the rst physical implementations of evolutionary robots 9, 1 7 ]. Despite the importance of keeping in mind the hard constraints of operating with physical robots, simulations still play an important role in evolutionary robotics for at least two sets of reasons:
doi:10.1007/3-540-64957-3_69 fatcat:jxrfotw7s5dqthj2gzunifmt7e