NEST-demoa co-simulation: Towards linking closed-loop motor control models to multi-scale experimental data
Project Description and Collaboration with HBP
In this project, an interface to couple two simulators, NEST and demoa, will be developed to enable co-simulation of spiking network models and musculoskeletal models taking into account both efferent and afferent communication signals. Technically, the co-simulation requires frequent MPI communication between the two simulators. An integration of the NEST-demoa interface with the interface solutions already under development within the HBP will be evaluated and similar design choices will be adopted if deemed appropriate. Such co-simulations will make it possible to extend and further strengthen analyses of brain activity at different scales (e.g., spikes and LFPs) in a “full neuronal cycle” including integrated environment dynamics through biophysical models. It will help in gaining more insights into the interplay of central nervous action and segment structure dynamics. The development of the interface will be use-case driven. On the demoa side, the use case under consideration represents the morphology of a primate arm. For this species, large data sets exist. The animal model will be based on morphometric data and consists of two arm segments: upper and lower arm, two joints: shoulder and elbow joint, and 21 muscles. The muscles will be model based on macroscopic muscle models. The data will be used for validation purposes and further analyses. On the NEST side, a suitable topdown model will be identified to assist in specifying the requirements. The main outcome of this project is the interface coupling the two simulators. As workflows for predicting LFPs (local field potentials) from NEST simulated networks (based on cosimulation with NEURON or Arbor) already exist, NEST-demoa co-simulation will also allow to link LFPs to MUPs (motor unit potentials). This means that data at different levels of observation can be linked (spiking activity and LFPs, single muscle fibers and MUPs).
The project NEST-demoa aims at interfacing demoa with NEST. NEST is a simulator for spiking neuronal networks offering its users great flexibility with respect to model definition. NEST is part of the Service Category SC3 for modelling and simulation workflows. NEST operates at the network description level, which means at the resolution of single neurons and synapses, where it is key that the interactions between neurons are realistic. It enables both bottom-up and top-down modelling approaches. Moreover, NEST is capable of exploiting the resources of modern supercomputers enabling brain-scale simulations, but it also performs well for small to medium sized network simulations. The design of coherent interfaces coupling NEST to other simulators, data analysis and visualization tools, and robotics applications is a major part of the development efforts in HBP SGA3. Co-simulations with other simulators operating at the cellular description level such as NEURON and Arbor (both part of SC3) facilitate predictions of mesoscopic measures such as local field potentials (LFPs) using the analysis tool LFPy. This allows for comparing not only spiking activity but also the resulting mesoscopic measures to experimental data, which makes it possible to link simulation studies to experimental findings at different scales of observation. Demoa is a simulator for biophysical human and animal neuromusculoskeletal models. Direct coupling of physical interaction with the environment to the neuronal state using mid to low level motor control models via biophysical muscle-tendon unit models is feasible. Multi-muscle system models allow for integrating direct computed feedback into brain-level control and thus, fully dynamic, neuro-mechanical systems. Predicted kinematics as well as neuronal states can be validated with experimental data, e.g. motion capture and EMG recordings. The use case for the work done in this project is the co-simulation of a primate (rhesus macaque) arm motion model using demoa and NEST.
Prof. Dr. Syn Schmitt,
Director, Institute for Modelling and Simulation of Biomechanical Systems
Professor for Computational Biophysics and Biorobotics at the University of Stuttgart
Syn Schmitt studied physics at the University of Stuttgart and graduated from the University of Tuebingen with a PhD in Theoretical Astrophysics (topic: muscle modelling / computational biophysics). In 2012, Schmitt was appointed as Juniorprofessor (assistant professor) at the University of Stuttgart. Since 2018, he is full professor of Computational Biophysics and Biorobotics at the University of Stuttgart and in 2019 he founded the Institute for Modelling and Simulation of Biomechanical Systems. Syn Schmitt is fellow of the Stuttgart Center for Simulation Science (SimTech) and a faculty member of the International Max Planck Research School for Intelligent Systems (IMPRS-IS). In 2019, he was appointed as Adjunct Professor in the School of Chemistry, Physics and Mechanical Engineering of the Queensland University of Technology in Brisbane, Australia. His research focusses on autonomous muscle-driven motion with special interests in design principles of the locomotion apparatus, non-linear dynamics of locomotion, motor control and morphological computation in biological and technical systems.
K. Ghazi-Zahedi, J. Rieffel, S. Schmitt, and H. Hauser. “Editorial: Recent Trends in Morphological Computation”. In: Frontiers in Robotics and AI 8 (2021), p. 159. DOI: 10.3389/frobt. 2021.708206.
M. Günther, R. Rockenfeller, T. Weihmann, D. F. B. Haeufle, S. Schmitt, and T. Götz. “Rules of nature’s Formula Run: Muscle mechanics during late stance is the key to explaining maximum running speed”. In: Journal of Theoretical Biology 523 (2021), p. 110714. DOI: 10. 1016/j.jtbi.2021.110714.
- J.R. Walter, M. Günther, D.F.B. Haeufle, and S. Schmitt.“A geometry- and muscle-based control architecture for synthesising biological movement”. In: Biological Cybernetics (2021). ISSN: 1432-0770. DOI: 10.1007/s00422-020-00856-4.
- D. F. B. Haeufle, I. Wochner, D. Holzmüller, D. Driess, M. Günther, and S. Schmitt. “Muscles reduce neuronal information load: quantification of control effort in biological vs robotic pointing and walking”. In: Frontiers in Robotics and AI – Soft Robotics 7.77 (2020). DOI: 10. 3389/frobt.2020.00077.
- I. Wochner, D. Driess, H. Zimmermann, D. F. B. Haeufle, M.Toussaint, and S. Schmitt. “Optimality Principles in Human Point-to-Manifold Reaching Accounting for Muscle Dynamics”. In: Frontiers in Computational Neuroscience 14 (2020), p. 38. DOI: 10.3389/fncom.2020. 00038.
- D. Driess, H. Zimmermann, S. Wolfen, D. Suissa, D. Haeufle, D. Hennes, M. Toussaint, and S. Schmitt. “Learning to Control Redundant Musculoskeletal Systems with Neural Networks and SQP: Exploiting Muscle Properties”. In: 2018 IEEE International Conference on Robotics and Automation (ICRA). 2018, pp. 6461–6468. DOI: 10.1109/ICRA.2018.8463160.
- D. F. B. Haeufle, B. Schmortte, H. Geyer, R. Müller, and S. Schmitt. “The Benefit of Com- bining Neuronal Feedback and Feed-Forward Control for Robustness in Step Down Per- turbations of Simulated Human Walking Depends on the Muscle Function”. In: Frontiers in Computational Neuroscience 12 (2018), p. 80. ISSN: 1662-5188. DOI: 10.3389/fncom.2018. 00080.