What is it about?
Locomotion is a based on the shape of a robot and the control of this shape, resulting in a gait. With soft robots you can explore a wide variety of shapes/gaits. Directly selecting on fitness explores only a part of the possible solutions, while with novelty search a greater diversity is possible, which is also benificial for the final fitness.
Featured Image
Photo by Arseny Togulev on Unsplash
Why is it important?
This is a text-book example of a case where novelty search outperforms traditional fitness–based search. Novelty search not only improved the performance and the diversity in the fitness space, but also contributed to a larger variety of possible solutions of locomotion on surfaces with different gravity.
Perspectives
Search is always focused on finding the optimal solution. The difference in performance between the different methods described in this paper seem to be subtle (but significant), but the locomotion strategies underlying this performance are really different. With novelty search one can see for instance tumbleweed, hoppers and 4-legged locomotion. As bonus also the subtle, but significant, benefits of two other textbooks methods are demonstrated on this case: the benefits of indirect encoding of shapes/gaits and the benefits of elitism when combined novelty search.
Dr. Arnoud Visser
Universiteit van Amsterdam
Read the Original
This page is a summary of: Novelty Search for Soft Robotic Space Exploration, July 2015, ACM (Association for Computing Machinery),
DOI: 10.1145/2739480.2754731.
You can read the full text:
Contributors
The following have contributed to this page
