The research field of our lab is the Sensorimotor Systems. Using experimental and computational approaches we test and refine theory behind the processing within neural structures to control sophisticated mechanical machinery of our body. If we steer true through the sea of complexity, we will eventually understand the brain by reverse engineering its functions with mathematical models. Our laboratory has multidisciplinary expertise: chronic long-term recording capabilities in animals with brain trauma, non-invasive neurophysiological research in humans, computational neuroscience and biomechanical research.

My strategy for the ‘reverse engineering’ of neural controller is to integrate the simplest computational models to aid in the interpretation of experimental data collected to test these same models and allow us to see the ‘big picture’. I am convinced that this framework will yield the highest degree of insight into the complex interactions within the neuro-musculo-skeletal system.

Our current research direction is focused on the principles of interactions between the mechanisms of neuromechanical hierarchy both in the context of stroke and spinal cord injury using animal models and in the context of improving control of advanced arm prosthesis for human amputees. One of the challenges for the current brain-machine interface is the lack of functional understanding of how neural processes interact within and across the different levels of neuraxis. Specifically, we have limited understanding of h ow cortical synergies or motor primitives are controlled to produce coupled sequential activation observed in reaching movements and locomotion. Lissencephalic (smooth) rat cortex is the perfect target for the microelectrode arrays with recording-stimulation capabilities to address this question. Building on my experience in recording and stimulation of cat motor cortex and brainstem structures I have collected preliminary data in rats using floating microelectrode arrays to demonstrate the feasibility of the methods. We have developed a new type of walkway specifically designed to create a dextrous locomotor task that requires cortical contribution in rodents. In addition, we are developing neuromechanical models for data processing that will guide our analysis.

My citation indices are h-index=16, i10index=18 with total number of citations at 1105. Complete List of Published Work in My Bibliography:

SPECIFIC PROJECTS

  • Musculoskeletal models for human-machine interface. The coordinated movements of a pianist’s hands or the precisely-timed swing of a batter are generated by muscles pulling on the high-dimensional and connected body segments. The recent development of advanced musculoskeletal models has contributed to the understanding of neural mechanisms or the nature of their disruption. The goal of the project is to develop validated computational tools that can quantify high dimensional whole-body movement using musculoskeletal (MS) dynamics optimized for real-time computations.
  • Spatiotemporal activity of the motoneurons (MN). The activation of spinal MNs represents the final neural output to the musculoskeletal system. The locomotor drive to lumbosacral MNs controlling hindlimbs in vertebrates is generated intrinsically in the spinal cord by interneuronal circuits capable of pattern generation even in the absence of supraspinal and peripheral inputs. In lower vertebrates (e.g. lamprey) locomotion is produced by a rostro-caudal wave of MN activity. Such a control strategy greatly simplifies the problem of controlling multiple degrees of freedom of the musculoskeletal system and suggests an attractive mechanism of the control of locomotion in higher vertebrates, e.g. rats and humans. (Yakovenko et al., 2002; Yakovenko et al., 2005).
  • Sensorimotor control of locomotion. Locomotion is generated by a hierarchically organized, neuro-musculo-skeletal (NMS) system with emergent properties that arise from dynamic interactions of its components. We study the holistic organization of the NMS in a series of experiments and computational models (Gillard, Yakovenko et al., 2000; Yakovenko et al., 2004; Prochazka & Yakovenko, 2007).
  • Restoration of locomotion after neural damage . Recent studies of long distance axon regeneration show improvements of motor functions observed in spinalized rats. Functional connections have been inferred from improvements in hindlimb locomotor movements even though regenerated neurons do not extend to the lumbosacral locomotor region. Using stimulation through microelectrode arrays, we can demonstrate that tonic intraspinal excitation immediately below a complete spinal cord transection improves motor performance in rats. This supports the hypothesis that nonspecific activation of thoracic propriospinal neurons can relay activation of the hindlimb locomotor networks (Yakovenko et al., 2007; Tuntevski et al., 2016).
  • Contribution of the motor cortex to the control of movement and posture . A fundamental question in motor control research is whether control of reaching evolved on the basis of the control of stereotypical movements like stepping over obstacles during locomotion. If so, similarity of the control of reaching and locomotion tasks should be reflected in conserved neural circuitry. In particular, the response of individual motor cortex neurons projecting to the spinal cord (PTNs) should be similar both during reaching to a lever while standing, and while stepping over obstacles during locomotion (Yakovenko & Drew, 2009; Yakovenko et al., 2011; Yakovenko & Drew, 2015).

Musculoskeletal Dynamics (MSD) for real-time limb control in VR/AR and prosthetics

18DoF biomimetic musculoskeletal model using direct synergy controller

  • Robust performance with low number of input signals controlling 29 muscles.
  • Minimal 'training' under 1 min. The training constitutes only the scaling of inputs (direct control).
  • Visual feedback to the subject is not necessary.

Acknowledgements

Developed in collaboration with Dr. Gritsenko, Mr. Boots, Mr. Sobinov, Dr. Fisher and Dr. Gaunt. Supported by DARPA HAPTIX award.

What happens if Stretch reflexes are dialed up too much?

Outtakes from Yakovenko et al., 2004 Biol. Cyb. The CPG drive is set to 25% of the level needed for load-bearing and the stretch reflexes are set to contribute 90%.


How about dialing up the CPG input to 50%!

How does activity of spinal cord look like?

Is this a wave? For details see Yakovenko et al., 2002.