Continuous then discrete: A recommendation for building robotic brains

Chris Eliasmith, P. Michael Furlong
Proceedings of the 5th Conference on Robot Learning, PMLR 164:1758-1763, 2022.

Abstract

Modern neural networks have allowed substantial advances in robotics, but these algorithms make implicit assumptions about the discretization of time. In this document we argue that there are benefits to be gained, especially in robotics, by designing learning algorithms that exist in continuous time, as well as state, and only later discretizing the algorithms for implementation on traditional computing models, or mapping them directly onto analog hardware. We survey four arguments to support this approach: That continuum representations provide a unified theory of functions for robotic systems; That many algorithms formulated as temporally continuous demonstrate anytime properties; That we can exploit temporal sparsity to effect energy efficiency in both traditional and analog hardware; and that these algorithms reflect the instantiations of intelligence that have evolved in organisms. Further, we present learning algorithms that are derived from continuous representations. Finally, we discuss robotic precedents for this approach, and conclude with the implications of using continuum representations in robotic systems.

Cite this Paper

BibTeX
@InProceedings{pmlr-v164-eliasmith22a, title = {Continuous then discrete: A recommendation for building robotic brains}, author = {Eliasmith, Chris and Furlong, P. Michael}, booktitle = {Proceedings of the 5th Conference on Robot Learning}, pages = {1758--1763}, year = {2022}, editor = {Faust, Aleksandra and Hsu, David and Neumann, Gerhard}, volume = {164}, series = {Proceedings of Machine Learning Research}, month = {08--11 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v164/eliasmith22a/eliasmith22a.pdf}, url = {https://proceedings.mlr.press/v164/eliasmith22a.html}, abstract = {Modern neural networks have allowed substantial advances in robotics, but these algorithms make implicit assumptions about the discretization of time. In this document we argue that there are benefits to be gained, especially in robotics, by designing learning algorithms that exist in continuous time, as well as state, and only later discretizing the algorithms for implementation on traditional computing models, or mapping them directly onto analog hardware. We survey four arguments to support this approach: That continuum representations provide a unified theory of functions for robotic systems; That many algorithms formulated as temporally continuous demonstrate anytime properties; That we can exploit temporal sparsity to effect energy efficiency in both traditional and analog hardware; and that these algorithms reflect the instantiations of intelligence that have evolved in organisms. Further, we present learning algorithms that are derived from continuous representations. Finally, we discuss robotic precedents for this approach, and conclude with the implications of using continuum representations in robotic systems.} }
Endnote
%0 Conference Paper %T Continuous then discrete: A recommendation for building robotic brains %A Chris Eliasmith %A P. Michael Furlong %B Proceedings of the 5th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2022 %E Aleksandra Faust %E David Hsu %E Gerhard Neumann %F pmlr-v164-eliasmith22a %I PMLR %P 1758--1763 %U https://proceedings.mlr.press/v164/eliasmith22a.html %V 164 %X Modern neural networks have allowed substantial advances in robotics, but these algorithms make implicit assumptions about the discretization of time. In this document we argue that there are benefits to be gained, especially in robotics, by designing learning algorithms that exist in continuous time, as well as state, and only later discretizing the algorithms for implementation on traditional computing models, or mapping them directly onto analog hardware. We survey four arguments to support this approach: That continuum representations provide a unified theory of functions for robotic systems; That many algorithms formulated as temporally continuous demonstrate anytime properties; That we can exploit temporal sparsity to effect energy efficiency in both traditional and analog hardware; and that these algorithms reflect the instantiations of intelligence that have evolved in organisms. Further, we present learning algorithms that are derived from continuous representations. Finally, we discuss robotic precedents for this approach, and conclude with the implications of using continuum representations in robotic systems.
APA
Eliasmith, C. & Furlong, P.M.. (2022). Continuous then discrete: A recommendation for building robotic brains. Proceedings of the 5th Conference on Robot Learning, in Proceedings of Machine Learning Research 164:1758-1763 Available from https://proceedings.mlr.press/v164/eliasmith22a.html.

Related Material