Abstract
Designing novel robots that can cope with a specific task is a challenging problem because of the enormous design space that involves both morphological structures and control mechanisms. To this end, we present a computational method for automating the design of modular robots. Our method employs a genetic algorithm to evolve robotic structures as an outer optimization, and it applies a reinforcement learning algorithm to each candidate structure to train its behavior and evaluate its potential learning ability as an inner optimization. The size of the design space is reduced significantly by evolving only the robotic structure and by performing behavioral optimization using a separate training algorithm compared to that when both the structure and behavior are evolved simultaneously. Mutual dependence between evolution and learning is achieved by regarding the mean cumulative rewards of a candidate structure in the reinforcement learning as its fitness in the genetic algorithm. Therefore, our method searches for prospective robotic structures that can potentially lead to near-optimal behaviors if trained sufficiently. We demonstrate the usefulness of our method through several effective design results that were automatically generated in the process of experimenting with actual modular robotics kit.
Highlights
The past several decades have witnessed a gradual increase in the use of robots for assisting humans in various activities from daily housekeeping to military service [1,2].Robot designs have been mostly restricted to a relatively narrow range of stereotypes such as robotic arms, wheeled vehicles, animal-like robots, or humanoids, despite the considerable diversity of tasks for which robots are used
Despite that the genetic evolution by itself contributes not much, combining it with the behavior optimization based on the reinforcement learning proves to be a highly effective method that produces highly competent designs, much better than the best one from ‘reinforcement learning of behaviors (R+L)’, and explores the promising area of the design space extensively. These results demonstrate that our method effectively reduces the design space based on the separation of structural and behavioral optimizations and efficiently finds prospective robotic structures that can potentially lead to near-optimal behaviors if trained sufficiently
We presented a method for discovering effective designs of modular robots using the genetic algorithm for evolving the structures of robots and applying the reinforcement learning algorithm to the evolved structures for training the behaviors of robots
Summary
The past several decades have witnessed a gradual increase in the use of robots for assisting humans in various activities from daily housekeeping to military service [1,2].Robot designs have been mostly restricted to a relatively narrow range of stereotypes such as robotic arms, wheeled vehicles, animal-like robots, or humanoids, despite the considerable diversity of tasks for which robots are used. Considering the large number of possible robot designs, it is highly probable that a more optimal design may exist for a given task; further, this design could be very different from stereotypical designs and may not be discovered via a manual approach. The problem of designing optimal robots with respect to a given task can be addressed computationally by applying an evolutionary algorithm to the population of virtual robots whose motions can be computed precisely in a physics-based simulator. Such a design approach–referred to as evolutionary robot design—has been studied extensively in the robotics and computer graphics communities [3]. In a typical method for evolutionary robot design, both the body morphology (structure) and the control mechanism (behavior) of each robot are encoded together as a single genotype
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
