Furman, Alex and Danielle, Nagar and Geoff, Nitschke (2019) Automating Collective Robotic System Design, Proceedings of IEEE Symposium Series on Computational Intelligence (IEEE SSCI 2019), 6-9 December 2019, Xiamen, China, IEEE.
![[img]](http://pubs.cs.uct.ac.za/style/images/fileicons/application_pdf.png) |
PDF
SSCI2019-v02.pdf Download (385kB) |
Abstract
This paper presents a study on methods for bodybrain (behavior-morphology) co-evolution in a collective evolutionary robotics system. We investigate a neuro-evolution developmental encoding method designed for the co-evolution of robot behavior-morphology couplings. This behavior-morphology evolution method is evaluated across increasingly complex (difficult) collective behavior task environments. This is in comparison to controller evolution within pre-engineered robot morphologies (sensory configurations). Task-complexity is equated with the degree of cooperation required in collective robotics tasks. Results indicate that the developmental method produces significantly more effective behavior-morphology couplings, compared to those evolved with direct encoding methods and controllers evolved within fixed morphologies. These results suggest that such developmental encoding methods could serve as a general evolutionary simulation design tool for automating collective robotic designs. An end goal is for such collective robotic system designs to be rapidly prototyped and deployed in the physical task environments for which they were evolved.
| Item Type: | Conference paper |
|---|---|
| Subjects: | Computing methodologies > Artificial intelligence |
| Date Deposited: | 20 Sep 2019 |
| Last Modified: | 10 Oct 2019 15:31 |
| URI: | http://pubs.cs.uct.ac.za/id/eprint/1322 |
Actions (login required)
 |
View Item |