Optimal Control of Multi-phase Movements with Learned Dynamics

Andreea Radulescu, Jun Nakanishi, Sethu Vijayakumar

Research output: Chapter in Book/Report/Conference proceedingConference contribution


In this paper, we extend our work on movement optimisation for variable stiffness actuation (VSA) with multiple phases and switching dynamics to incorporate scenarios with incomplete, complex or hard to model robot dynamics. By incorporating a locally weighted nonparametric learning method to model the discrepancies in the system dynamics, we formulate an online adaptation scheme capable of systematically improving the multi-phase plans (stiffness modulation and torques) and switching instances while improving the dynamics model on the fly. This is demonstrated on a realistic model of a VSA brachiating system with excellent adaptation results.
Original languageEnglish
Title of host publicationMan–Machine Interactions 4
Subtitle of host publication4th International Conference on Man–Machine Interactions, ICMMI 2015 Kocierz Pass, Poland, October 6–9, 2015
EditorsAleksandra Gruca, Agnieszka Brachman, Stanislaw Kozielski, Tadeusz Czachórski
Place of PublicationCham
PublisherSpringer International Publishing
Number of pages16
ISBN (Electronic)978-3-319-23437-3
ISBN (Print)978-3-319-23436-6
Publication statusPublished - 9 Sep 2016
Event4th International Conference on Man–Machine Interactions - Kocierz Pass, Poland
Duration: 6 Oct 20159 Oct 2015

Publication series

NameAdvances in Intelligent Systems and Computing
PublisherSpringer International Publishing
ISSN (Print)2194-5357


Conference4th International Conference on Man–Machine Interactions
Abbreviated titleICMMI 2015
CityKocierz Pass


Dive into the research topics of 'Optimal Control of Multi-phase Movements with Learned Dynamics'. Together they form a unique fingerprint.

Cite this