Evolved Transistor Array Robot Controllers¶
Publication: Evolutionary Computation, Vol. 28, No. 4, pp. 677–708, MIT Press, 2020
Date: 2020
Abstract
This paper represents the first use of a field programmable transistor array (FPTA) to develop robot control circuits through direct evolution in analog hardware. The team successfully created controllers for a physical robot capable of executing multiple visually-guided tasks, including locating targets within complex settings. The system also evolved circuits for speech recognition, enabling voice-activated behavioral switching. A notable innovation was the coevolution of visual features alongside controllers, accomplishing simultaneous dimensionality reduction, feature extraction, and selection. The work demonstrates that intrinsic analog evolution remains a viable approach for producing sophisticated robot controllers.
Suggested Citation
Garvie, M., Flascher, I., Philippides, A., Thompson, A., & Husbands, P. (2020). Evolved transistor array robot controllers. Evolutionary Computation, 28(4), 677–708. MIT Press.
Authors¶
- Michael Garvie
- Ittai Flascher
- Andrew Philippides
- Adrian Thompson
- Phil Husbands