Safe Robot Learning in Assistive Devices through Neural Network Repair

Keyvan Majd, Geoffrey Mitchell Clark, Tanmay Khandait, Siyu Zhou, Sriram Sankaranarayanan, Georgios Fainekos, Heni Amor
Proceedings of The 6th Conference on Robot Learning, PMLR 205:2148-2158, 2023.

Abstract

Assistive robotic devices are a particularly promising field of application for neural networks (NN) due to the need for personalization and hard-to-model human-machine interaction dynamics. However, NN based estimators and controllers may produce potentially unsafe outputs over previously unseen data points. In this paper, we introduce an algorithm for updating NN control policies to satisfy a given set of formal safety constraints, while also optimizing the original loss function. Given a set of mixed-integer linear constraints, we define the NN repair problem as a Mixed Integer Quadratic Program (MIQP). In extensive experiments, we demonstrate the efficacy of our repair method in generating safe policies for a lower-leg prosthesis.

Cite this Paper

BibTeX
@InProceedings{pmlr-v205-majd23a, title = {Safe Robot Learning in Assistive Devices through Neural Network Repair}, author = {Majd, Keyvan and Clark, Geoffrey Mitchell and Khandait, Tanmay and Zhou, Siyu and Sankaranarayanan, Sriram and Fainekos, Georgios and Amor, Heni}, booktitle = {Proceedings of The 6th Conference on Robot Learning}, pages = {2148--2158}, year = {2023}, editor = {Liu, Karen and Kulic, Dana and Ichnowski, Jeff}, volume = {205}, series = {Proceedings of Machine Learning Research}, month = {14--18 Dec}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v205/majd23a/majd23a.pdf}, url = {https://proceedings.mlr.press/v205/majd23a.html}, abstract = {Assistive robotic devices are a particularly promising field of application for neural networks (NN) due to the need for personalization and hard-to-model human-machine interaction dynamics. However, NN based estimators and controllers may produce potentially unsafe outputs over previously unseen data points. In this paper, we introduce an algorithm for updating NN control policies to satisfy a given set of formal safety constraints, while also optimizing the original loss function. Given a set of mixed-integer linear constraints, we define the NN repair problem as a Mixed Integer Quadratic Program (MIQP). In extensive experiments, we demonstrate the efficacy of our repair method in generating safe policies for a lower-leg prosthesis.} }
Endnote
%0 Conference Paper %T Safe Robot Learning in Assistive Devices through Neural Network Repair %A Keyvan Majd %A Geoffrey Mitchell Clark %A Tanmay Khandait %A Siyu Zhou %A Sriram Sankaranarayanan %A Georgios Fainekos %A Heni Amor %B Proceedings of The 6th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2023 %E Karen Liu %E Dana Kulic %E Jeff Ichnowski %F pmlr-v205-majd23a %I PMLR %P 2148--2158 %U https://proceedings.mlr.press/v205/majd23a.html %V 205 %X Assistive robotic devices are a particularly promising field of application for neural networks (NN) due to the need for personalization and hard-to-model human-machine interaction dynamics. However, NN based estimators and controllers may produce potentially unsafe outputs over previously unseen data points. In this paper, we introduce an algorithm for updating NN control policies to satisfy a given set of formal safety constraints, while also optimizing the original loss function. Given a set of mixed-integer linear constraints, we define the NN repair problem as a Mixed Integer Quadratic Program (MIQP). In extensive experiments, we demonstrate the efficacy of our repair method in generating safe policies for a lower-leg prosthesis.
APA
Majd, K., Clark, G.M., Khandait, T., Zhou, S., Sankaranarayanan, S., Fainekos, G. & Amor, H.. (2023). Safe Robot Learning in Assistive Devices through Neural Network Repair. Proceedings of The 6th Conference on Robot Learning, in Proceedings of Machine Learning Research 205:2148-2158 Available from https://proceedings.mlr.press/v205/majd23a.html.

Related Material