Using Inertial measurements to quantify and classify movements for health care analysis.

W.S.Harwin

http://www.reading.ac.uk/~shshawin

21 March 2018

The ability to control and maintain force between robots and people remains a major problem in assistive robotics. This talk will review some past work in assistive robotics and machine mediated stroke rehabilitation as well as explore the challenges in designing effective force control for rehabilitation applications.

Haptics

Machine mediated rehabilitation

Robot equations

\[ \underline\tau= M(\underline{q})\ddot{\underline{q}}+C(\underline{q},\dot{\underline{q}})\dot{\underline{q}} +G(\underline{q}) \]

i.e. \[ F=m\ddot{x} \]

or

\[ \ddot{x}=\frac{F}{m} \]

http://www.cybernetia.co.uk/LN/Inertia.html An adventure game with gravity

Accelerations of lower leg

\[~^1\dot{\vec\omega}_1=\begin{bmatrix} 0\\ 0\\ {\ddot\theta}_{1} \end{bmatrix}\] \[~^1\dot{\vec{v}}_{cog_1}=\begin{bmatrix} g\,\sin\left(\theta _{1}\right)-{{\dot\theta}_{1}}^2\,\mathrm{lcog}_{1}\\ {\ddot\theta}_{1}\,\mathrm{lcog}_{1}+g\,\cos\left(\theta _{1}\right)\\ 0 \end{bmatrix}\]

Accelerations of the thigh

\[~^2\dot{\vec\omega}_2=\begin{bmatrix} 0\\ 0\\ {\ddot\theta}_{1}+{\ddot\theta}_{2} \end{bmatrix}\] \[~^2\dot{\vec{v}}_{cog_2}=\begin{bmatrix} \sin\left(\theta _{2}\right)\,\left({\ddot\theta}_{1}\,l_{1}+g\,\cos\left(\theta _{1}\right)\right)-\mathrm{lcog}_{2}\,{\left({\dot\theta}_{1}+{\dot\theta}_{2}\right)}^2-\cos\left(\theta _{2}\right)\,\left({{\dot\theta}_{1}}^2\,l_{1}-g\,\sin\left(\theta _{1}\right)\right)\\ \cos\left(\theta _{2}\right)\,\left({\ddot\theta}_{1}\,l_{1}+g\,\cos\left(\theta _{1}\right)\right)+\sin\left(\theta _{2}\right)\,\left({{\dot\theta}_{1}}^2\,l_{1}-g\,\sin\left(\theta _{1}\right)\right)+\mathrm{lcog}_{2}\,\left({\ddot\theta}_{1}+{\ddot\theta}_{2}\right)\\ 0 \end{bmatrix}\]

Accelerations of the trunk

\[~^3\dot{\vec\omega}_3=\begin{bmatrix} 0\\ 0\\ {\ddot\theta}_{1}+{\ddot\theta}_{2}+{\ddot\theta}_{3} \end{bmatrix}\] \[~^3\dot{\vec{v}}_{cog_3}=\begin{bmatrix} \sin\left(\theta _{3}\right)\,\left({\ddot\theta}_{1}\,l_{2}+\cos\left(\theta _{2}\right)\,\left({\ddot\theta}_{1}\,l_{1}+g\,\cos\left(\theta _{1}\right)\right)+\sin\left(\theta _{2}\right)\,\left({{\dot\theta}_{1}}^2\,l_{1}-g\,\sin\left(\theta _{1}\right)\right)\right)-\cos\left(\theta _{3}\right)\,\left(l_{2}\,{\left({\dot\theta}_{1}+{\dot\theta}_{2}\right)}^2-\sin\left(\theta _{2}\right)\,\left({\ddot\theta}_{1}\,l_{1}+g\,\cos\left(\theta _{1}\right)\right)+\cos\left(\theta _{2}\right)\,\left({{\dot\theta}_{1}}^2\,l_{1}-g\,\sin\left(\theta _{1}\right)\right)\right)-\mathrm{lcog}_{3}\,{\left({\dot\theta}_{1}+{\dot\theta}_{2}+{\dot\theta}_{3}\right)}^2\\ \cos\left(\theta _{3}\right)\,\left({\ddot\theta}_{1}\,l_{2}+\cos\left(\theta _{2}\right)\,\left({\ddot\theta}_{1}\,l_{1}+g\,\cos\left(\theta _{1}\right)\right)+\sin\left(\theta _{2}\right)\,\left({{\dot\theta}_{1}}^2\,l_{1}-g\,\sin\left(\theta _{1}\right)\right)\right)+\sin\left(\theta _{3}\right)\,\left(l_{2}\,{\left({\dot\theta}_{1}+{\dot\theta}_{2}\right)}^2-\sin\left(\theta _{2}\right)\,\left({\ddot\theta}_{1}\,l_{1}+g\,\cos\left(\theta _{1}\right)\right)+\cos\left(\theta _{2}\right)\,\left({{\dot\theta}_{1}}^2\,l_{1}-g\,\sin\left(\theta _{1}\right)\right)\right)+\mathrm{lcog}_{3}\,\left({\ddot\theta}_{1}+{\ddot\theta}_{2}+{\ddot\theta}_{3}\right)\\ 0 \end{bmatrix}\]

Sphere:

Sensor Platform for HEalth in a Residential Environment

  • Funded by the EPSRC
  • University of Bristol
  • University of Southampton
  • University of Reading
  • Knowle West Media Centre
  • Toshiba
  • Bristol City Council
  • IBM

  • Video sensing
    • Activity monitoring
    • Social interaction
    • Movement about the home
  • Wearable technology
    • Activity monitoring
    • Quality of movement
  • Environmental sensors
    • Temperature, light, air quality, humidity
    • Water and electricity consumption

Move to PPT

Sit to stand

  • Video shows Baysian classifier of sittting, standing sit-to-stand and stand-to-sit
  • Left classifier uses acceleration mean and variance
  • Right classifier uses angular velocity

Sit to stand Study

  • Data gathering in the lab and in unstructured environments
  • Data sets that include
    • young adults
    • older adults
    • people with Parkinson's disease
  • Data collected (and maintained) by Southampton University
  • Sensors developed by Reading University

Acknowledgements

Ian Craddock
Margaret Cox
Barry Quinn
William Hoderbaum
Simon Sherratt
Faustina Hwang
Brian Tse
Emma Villeneuve
Farshid Amirabdollahian
Rui Loureiro
Ally Barrow
Balazs Janko
Rachel King
Ozan Tokatli
Maitreyee Wairagkar
Colleagues at Royal Berkshire Hospital, Southampton University, Bristol University and Kings College, London