LEVIN

Home About Me Medical Robotics Robotic Mobility Other Projects CV Fun & Volunteer

Robotic Mobility

I have studied robotic mobility for capsule endoscopes (or colonoscopes) from two main perspectives. The first perspective examines the robotic capsule endoscope mobility system contact problem for understanding how to optimize traction generation. The second perspective examines the robotic capsule endoscope housing contact problem for predicting how much traction is needed to overcome frictional forces and mobilize the capsule endoscope through the gastrointestinal tract.


Robotic mobility system contact problem

The first perspective examines the robotic capsule endoscope mobility system contact problem. It is important to examine the contact problem of the interaction between the robotic locomotion system and the interfacing tissue so that future tread designs can be optimized. For the capsule endoscope that we have developed, micro-patterned polydimethylsiloxane (PDMS) treads (or wheels) have been used as the mobility method due to its friction enhancing properties. This contact problem can be broken into two perspectives: 1) a global perspective where we examine the wheel-tissue interface, and 2) a local perspective where we examine the interaction between individual micro-pillars and the underlying tissue. I have studied this problem experimentally and by using models (empirical, finite element and analytical).



Robotic mobility on biological tissue contact problem



The robotic mobility contact problem, consisting of a micro-patterned PDMS robotic wheel on a biological substrate, is multi-faceted. To simplify the problem, it is examined from two different perspectives: 1) a global perspective (figure, left) which involves the wheel and the underlying tissue substrate, and 2) a local perspective (figure, right) which isolates a single micro-pillar and its interaction with the underlying substrate.


The problem is multi-faceted due to the complex underlying physics. Contributing forces in a free body diagram include, but are not limited to, friction, van der Waals, capillary, adhesion, and forces from mechanical engagement. The wheel and micro-patterned PDMS tread are assumed to be rigid compared to the soft biological tissue. This is an appropriate assumption as the elastic modulus of the PDMS is at least one order of magnitude larger than the elastic modulus of the tissue. The real problem involves a lubrication layer (e.g., mucus), but this is disregarded for most of the experiments and modeling.


The robotic wheel has  rotational speed, translational speed, and normal force. The micro-pattern has an elastic modulus, size, and geometry. All of these factors contribute to the traction force generated by a wheel/tread.



Micro-patterns for friction enhancement



The technique of using micro-patterns for friction enhancement stems from biology, where microscopic structures (called setae) are found on the end of beetle feet (similar structures are also found on the pads of gecko feet). These micro-structures enhance friction for the insect, enabling them to walk up walks and across ceilings. These micro-patterns can be recreated on the surface of a polymer, using a micro-mold, fabricated by a photolithography process. My dissertation studied the effect of micro-pattern geometry, size and PDMS modulus.


Journal of Medical Devices article

Experimental evaluation of robotic mobility



To evaluate the traction force generation of a micro-patterned wheel on a soft substrate, an automated traction measurement (ATM) platform for developed. The device consists of a load platform which measures normal and transverse forces, a linear slider assembly to induce translational motion, a passively pivoting horizontal arm, an actuator to induce rotational motion, and an actuated counterweight to control the normal force of the wheel. Closed-loop control was implemented for translational speed, rotational speed, and normal force.


The ATM platform was used to t study the effects of translational speed, slip ratio, normal force, micro-pattern size, micro-pattern pillar geometry, substrate stiffness and tread modulus. The device was also used to validate finite element and analytical models.


Transactions on Mechatronics article
Transactions on Biomedical Engineering

Finite element modeling to predict traction force



A finite element model was developed to predict traction force of a micro-patterned wheel. The model consisted of a soft (viscoelastic and hyperelastic) substrate, a rigid wheel (both smooth and micro-patterned), and a velocity dependent coefficient of friction.


To increase repeatability, a synthetic tissue was used. The synthetic tissue was characterized with respect to viscoelasticity, hyperelasticity and Poisson's ratio.



Finite element and analytical modeling results



The finite element model results were validated using both experimental and an analytical solution. The model and empirical data matched well with respect to traction force, differential velocity, normal force, and global deflection.


Robotic capsule endoscope housing contact problem

The second perspective examines the robotic capsule endoscope housing contact problem. It is important to examine the interaction between the housing and the surrounding gastrointestinal (GI) tissue to determine how much traction force is needed to overcome frictional (and other) forces on the capsule housing with an end goal of locomoting the capsule through the GI tract. This work was performed at the BioRobotics Institute of Scuola Superiore Sant'Anna in Pisa, Italy.



Capsule endoscope housing contact problem definition



The contact problem for the capsule endoscope housing was developed with a magnetic locomotion system in mind (figure, top), but could be applied for any locomotion system. In a typical magnetic locomotion scenario, the capsule has an embedded internal permanent magnet (IPM) that is attracted to an external permanent magnet (EPM). The EPM is fixed to the end of a robotic arm that controls the EPM's location and orientation. Changes in location and orientation of the EPM results in changes in location ad orientation of the IPM and therefore the capsule endoscope. To simplify this problem so that a model could be developed, an equivalent benchtop diagram was developed (figure, bottom). The model assumes contact with only one side of the capsule (e.g., in the case of an insufflated colon). The benchtop experimental setup was derived from the magnetic scenario, with an adjusted capsule mockup weight that accounted for the magnetic attraction force.



Analytical model development



The model was developed assuming that the capsule sat on a tissue substrate of height H. The tissue was assumed to be viscoelastic and conform to the outer surface of the capsule. The shape function for the underlying tissue modeled haustral folds as a cosine function with height (amplitude) h and width (quarter period) w. The capsule had dimensions D (diameter), R (edge radius), and L (length), weight FW, and constant speed v.


The model consisted of several controllable inputs (capsule dimensions, weight and speed) as well as several uncontrollable but measurable inputs (tissue properties and surface topography).



Experimental validation



In order to validate the model empirically, several experiments and capsule mockups were developed. Baseline parameters were defined for capsule dimensions, weight and speed based on the commercially available PillCam Colon. These parameters were varied to create a range of values. Capsule mockups were fabricated (3D printed) with varying dimensions and weights.


Drag tests were performed on smooth excised tissue samples using a light-weight capsule to measure the coefficient of friction. Drag tests were performed on random tissue samples using all of the capsule to measure a maximum drag force, which was then used to validate the model.


Viscoelastic tissue parameters for each tisue sample were determined by curve fitting a Double Maxwell-arm Wiechert (DMW) model to experimental data from stress-relaxation indentation tests.


Tissue topography parameters (h and w) were determined by scanning the surface of each tissue sample using an interferometer to generate a 3D map. Slices of the 3D map were analyzed to calculate averate peak-to-peak amplitude and periods.



Model and experimental validation results



There was an average normalized root mean square error (NRMSE) of 6.25% between all experimental data and analytical model output. The model was validated with respect to capsule diameter (D), capsule edge radius (R), capsule length (L), capsule speed (v), capsule weight (FW), and GI tract region (esophagus, stomach, small intestine and large intestine). By taking tissue samples from all regions of the GI tract, we were able to effectively validate the model with respect to tissue properties and tissue surface morphology.


LEVIN SLIKER | PHD MECHANICAL ENGINEERING


7801 Valmont Rd, Boulder, CO 80301 | Tel: 719 460 3866 | levin.sliker@gmail.com