CPS: Medium: Programmable Second Skin to Re-educate Injured Nervous Systems
Objectives and approaches.
The goal of this project is to create a novel cyber-physical system (CPS), a self-configuring second skin, consisting of soft programmable materials. The second skin is worn on the skin surface over affected body parts of individuals who have suffered brain injury. The second skin may promote restoration of function in brain-injured infants by using its miniaturized sensors and actuators to expand their restricted range of exploratory behaviors. Each specific aim below follows from principles of biologically-inspired engineering that we are using in the design, test, and implementation of this new approach to rehabilitating injured nervous systems.
A.Specific Aim 1. Microfabrication of second skin programmable material.
Figure 1 shows the design and the performance of the artificial skin for multi-modal sensing capability that detects multi-axial strain and contact pressure. Micro-channels, filled with a liquid metal alloy, Eutectic Gallium Indium (EGaIn), are embedded in an elastomer sheet. When the material experiences strain in the axial direction of the channels, the overall channel length increases and the cross-sectional areas of the channels decrease, which causes an increase in the overall channel resistance. Also, when the surface is compressed the sensor can detect the applied pressure. The sensor has capability of decoupling different types of stimuli.
Each miniature pneumatic actuator, a synthetic muscle, is composed of an inextensible thread wrapped around a rubber cylinder whose modulus is calculated in Figure 2 at left. Filling the actuator with a pressure P>0 will cause the actuator to shorten. We fabricated functional units consisting of a miniaturized pneumatic muscle and a hyperelastic strain sensor in successive layers. Filling each actuator with compressed air resulted in a length change that constituted a feedback loop for detecting synthetic muscle contraction.
B.Specific Aim 2: Achieving a new understanding of the development of early sensorimotor control.
Exploratory movements serve to register the surrounding bath of force fields and so to engender a flow of information from the environment to the more central structures regulating organism behavior. The exploratory coupling between organism and environment must occasion just such a distal-to-proximal flow where changes at the motor periphery stimulate later changes in the central regulatory structures of the motor system. We sought to probe the dynamics of infant spontaneous kicking for evidence of this distal-to-proximal flow of information. Documenting the flow of information requires identifying empirical markers of how well organisms detect information. Recent work has suggested that the detection of information for perception-action may be indexed by multifractal fluctuations in exploratory movements. That is, fluctuations composing exploratory movements exhibit a variety of fractal scaling exponents, and this rich variety of fractal scaling exponents predict the changes in the use of contextual energy distributions for perceptual judgments, both across time and across participant. Multifractal fluctuations during exploration of the task environment appear to mirror multifractal fluctuations available in the task environment. Furthermore, this mirroring is not simply of the environmental fluctuations "on average," that is, the additive structure (i.e., mean, variance, and autocorrelation but it rather reflects sensitivity to the multifractality attributable to multiplicativity in the fluctuations, above and beyond the aggregate, additive structure. That is to say, it may be possible to identify flows of information by examining flows of multiplicative multifractal fluctuations and thus to bring new rigor to tests of the exploratory aspect of spontaneous movements.
We sought to bring these notions to bear on infant spontaneous kicking. That is, if exploration at the scale of the organism-environment relationship could be considered the propagation of multiplicative multifractal fluctuations, it is possible that spontaneous kicking might exhibit similar propagation of multiplicative multifractal fluctuations. Whereas previous work had only considered the flow of information between organism and task environment, we were curious whether similar statistical relationships might be found among components of the motor system, that is, between ankle and knee, between knee and hip. Perhaps the activity of different joints along the leg might show a similar pattern of multifractal fluctuations spreading from one to the other. If the pattern of multifractal fluctuations reveals effects of multiplicativity from more distal joints (e.g., ankle) to more proximal joints (e.g., hip), this evidence would confirm the exploratory aspect of spontaneous kicking in human infants. In other words, if relatively proximal joints absorb the multiplicative fluctuations of relatively distal joints, then we begin to see specific evidence of spontaneous kicking movements ferrying information about the surrounding force-field bath towards the central nervous system.
We used motion capture of 5 infants each tested at 3, 4.5, and 6 months of age, to generate joint-angle time series for the ankle, the hip, and the knee for several consecutive 30-second intervals. We computed the multifractality for each original time series as well as for fifty surrogate time series each mimicking the aggregate, additive structure. We scaled the original series' multifractality by the average multifractality of the surrogates to derive a ratio expressing the degree of multiplicativity. We modeled how this multiplicativity ratio changed across the three joints for each leg using vector autoregression (VAR), a modeling strategy used to assess mutual effects among interacting, bidirectionally coupled variables VAR modeling both controls for the autoregressive behavior of each individual variable while shedding light on the unique effects of each variable's current behavior on each other variable's later behavior. VAR modeling showed that there were distal-to-proximal flows of multiplicative multifractal fluctuations on both legs, as evidenced by positive impulse responses in multiplicativity ratios from ankle to knee, from ankle to hip, and from knee to hip (Figure 2). Thus, we found evidence that multiplicative multifractal fluctuations at the relatively distal joints engender multiplicative multifractal fluctuations at relatively proximal joints.
C.Specific Aim 3. Developing new bio-inspired programming techniques for encoding a network of distributed actuators and sensors to robustly cooperate with the human body and with each other during task performance.
A challenge in modular design is to use different combinations of sensor-actuator units to achieve specific collective behaviors, such as simultaneous contraction, sequential contraction, or bending. The architecture of a hollow elastomeric cylinder (Figure 4) was used to conduct experiments that demonstrated the collective actuation of 16 sensor-actuator units. An example of contraction is shown.
The system software architecture is divided into two main layers: System Services Layer (SSL), and Application Layer (AL) (see Figure 9). The former implements fundamental components that manage local resources and provide primitives to support algorithms at the AL. The SSL implements a clock-driven scheduler, handles inter-module communication with neighboring modules, accesses strain sensor readings, and sets actuation parameters. The AL specifies the application goal using the services provided by the SSL. The clock-driven scheduling provides predictable execution of specific tasks at individual modular units, allowing simplification of control on timing of sensing, processing, and actuation tasks. When multiple muscles are actuated (contracted) collectively as a group, there is an overall displacement and force produced.