Design of a Terrain Detection System for Foot Drop

PowerPoint Presentation

Design of a Terrain Detection System for Foot Drop Christopher R. SullivanMechanical EngineeringOctober 25th, 2012Name, department, event, date, advisorProject GoalCreate an Ankle mounted system for identifying specific ground conditionsTodays talkBackgroundLiterature reviewPendulum Analysis of GaitExperimental MethodsTerrain CharacterizationPattern RecognitionConclusionsFuture Work

Image Source: http://transit-safety.volpe.dot.gov/publications/safety/pedestrian/html/images/dot-tsc-umta-84-36_p0009a.gifMy goal is to create a system for recognizing changes in ground conditions. It should be contained within the AFO, which means it must be mounted to the ankle. In todays talk we will go over what any of that means and form there how this problem was discovered, and steps that I have taken in solving the problem

PER-A-Nee-il

2BackgroundWhat is Foot drop?It is a symptomInability/ weakness of the ankle

Image Source: http://sports.jrank.org/article_images/sports.jrank.org/dorsiflexion.1.jpgFoot Drop is characterized by Inability to control the foot during the Gait CycleStrokeCerebral PalsyDirect injury to the Peroneal Nerve

3Gait Cycle and Foot Drop100%Foot StrikeOpposite Toe-OffOpposite Foot StrikeToe-OffFoot ClearanceTibia VerticalFoot StrikePeriodsInitial Double-Limb SupportSingle-Limb StanceSecond Double-Limb SupportInitial SwingMid-SwingTerminal SwingStanceSwing% of Cycle0%62%PeriodsInitial SwingMid-SwingTerminal SwingToe-OffSwing% of Cycle62%100%Mid TripFoot CrashFallenFoot Drop is characterized by Inability to control the foot during the Gait CycleStrokeCerebral PalsyDirect injury to the Peroneal Nerve

4Ankle Foot OrthoticReplace lost ankle functionalityCorrect brace for the correct problemWide variety

Image Source: http://www.mobilelimbandbrace.com/images/Articulating_AFO_Overlap.gifhttp://www.spsco.com/assets/images/dynamic-walk-single-side-2_large.jpghttp://proactiveasia.com/web_image/orthotics/Trulife%20semi-solid-afo_web.gif

Jointed BraceDynamic Walk

Solid BraceCurrently aofs are the leading for of brace for combating foot drop. The work by limiting the maximum plantar flexion that a patients foot can achievePateints and AFOs varry widly based on the sivarity of a patients disability, foot drop can be compounded with toe currling and rolling of the ankle, all of which would require a studiur braceThe problem with this is there are times when allowing more flexsion of the ankle is actually benificial to the patient, such as when someone is walking down a ramp of down a set of stairs.

5Stakeholder InterviewsInterviewed Clients, Clinicians,and Prosthetist OrthotistClientele QuestionDo you have any specific complaints about your AFO?Do you have any specific compliments about your AFO?How many AFOs have you had?How long have any of your AFOs lasted?What kind of hinges have your past or present AFOs had?If you could remove material from your AFO, where would you remove it from?

Image Source: http://www.humanresourcesdegree.net/images/stories/School%20Logos%20-%20Masters/NazarethCollege.jpg http://www.workforcediversitynetwork.com/images/logos/RGHS_stacked_150.jpghttp://www.rochesterorthopedic.com/

The very first thing I did was to complete patient interveiws, to determin if there was any problem that they all seemed to face.It is important to keep in mind the wide vareations within the pateint population, yet they all seem to feel instability while walking down stairs and ramps.This is because as I have stated earlier their brace is hindering them from making a proper contact with the ground.All of the patients complained about the instability they felt on stairs

Explain why they feel uncomfortable !!!!!!!!!Rochester orthopedic PEDIC LABS

Peoples needs vary widely!Major Interview TakeawaysFoot drop has many other compound symptomsProsAllow clients to walkConsweight/bulk of the AFOInstability on ramps and stairsAFO users needs can differ widely

Device BasicsFunctionalityProvide appropriate support for the foot at the appropriate timeStairs RampLevel surfaceProvide accurate assessment of ground conditions, before the foot hits the ground.Splitting the projectBraceControl System

These are just a few quick customer needs that I have defined for the overall project. This includes work that will be done by a senior design team to actually building an afo to take advantage of my system.8Literature ReviewCommercially available AFOsExperimental AFOsHuman Gait AnalysisTerrain Detection

Commercially Available AFOsHard Plastic BraceNo JointJointedTamarackE-StimBio-nessWalkAide

Image Source: http://www.orthomedics.us/SiteImages/Bioness%20orange.gif http://www.orthomedics.us/SiteImages/WalkaideCuffGray.gifExperimental AFOsI-AFOAir Muscle AFOPneumatic Power Harvesting AFO

AFOMetricSpeci-AFOMaximum braking torque10 (Nm)Mass990 (g)Movable angle45 to +45 (deg)Air Muscle AFOMaximum pulling torque171.7 (Nm)Mass1.3-1.7 (kg)Movable angle10 to +35 (deg)Pneumatic Power Harvesting AFOMaximum power generation10(W)Mass1 (kg)Movable angle9 to +15 (deg) i-AFO used a rotational braking system to variably dampen the systemAir Muscle AFO.Pneumatic Power Harvesting AFO This AFO used a very small piston to mechanically lock the foot into a preferable position11Human Gait Analysis2-D Modeling of the ankle foot systemLimitationsEquipmentCoordinate SystemsAccelerometers & Gyroscopes

Image Source: http://physio.otago.ac.nz/images/clinics/gait2.jpgTerrain DetectionRoboticsLasers GPSObject Avoidance

Image Source: http://www.pbs.org/wgbh/nova/darpa/images/cars-03-stanley-image3-l.jpgThe current path that terrain detection is taking is much like human gait analysis: As time goes on, the field has become more complicated and more accurate, but little thought has been given to simple classification of different types of terrain into broad families.13Gaps in LiteratureLack of design possessAdaptive AFOsPendulum Analysis of GaitMotivationAssumptionsMethodsLaGrange's Methody12345m4m5m3m2m1xLaGrange's MethodResultsDynamic Analysis of MomentQuasistatic Analysis of Moment Experimental MethodExperiments?Devices UsedPMD-1208LSSharp GP2Y0A02YKPiezo electric plateSharp GP2D12

Experiment #First SensorSecond SensorGround Type IdentificationPredictions SuccessfulRun 1IR: GP2Y0A02YKN/AYesN/ARun 2IR:GP2Y0A02YKPiezo electric plateYesNoRun 3IR:GP2Y0A02YKIR:GP2D12YesYesWalking ScenariosLevel WalkingUp and down stairsUp and down rampsLong walkLevel walking: Recorded for approximately 20 seconds. Avoided walking right next to walls or chair legs, as these objects might be picked up by the sensor.Up and down stairs: Recorded a flight of stairs, ensuring that the leg without the device leads, as this will likely be the case for someone with an injured leg.Up and down ramps: Recorded for approximately 20 seconds up or down a ramp.Long walk, multiple terrain types to be differentiated (e.g., down ramp, up ramp, level, upstairs, level).

19Terrain CharacterizationImportance?Fourier SeriesTime ShiftingRANSACExperiments?Core of this project is the ability to reliably generate a model for a given set of data to be used later for determining what kind of ground is the client walking over20Fourier SeriesHarmonicsPeriod scalingImportance of order

Order4th order selected

Time ShiftingAngle Addition Formula Time Shift (t- t0)Scaled by Gait Speed (w)RANSACRANdom SAmple ConsensusStarting AssumptionsInliersFit to a modelOutlinersCan come from noisy data, or erroneous assumptions

Image Source: http://en.wikipedia.org/wiki/RANSACRANSACInput:DataMinimum number of data pointsAdded guess at the gait periodNumber of iterationsMinimum errorMinimum number of points for a real modelOutput:Best model Best consensus set Best errorCoefficientSpecTolerance19%Min Random # of Points55% of # of points in a stepMin Points for Model55% of # of points in data setNumber of Iterations300Determining Gait Period Run #1Manually Guessing Gait SpeedNo Algorithm UsedDetermining Gait Period Run #2Piezoelectric plateIdentifying the proper gait periodPick two similar high point and test the rest of the data for similaritiesDifferentiation between direction change and heel strike

Level Walking Gait Period AnalysisWhen the results are not so pretty

Down Stairs Gait Period AnalysisImprovements?Foot not striking in the location of the plateMultiple SensorsFilter

Determining Gait Period Run #3

Time shift of .03 run time .003Inliers outliners mean STD30Logic LoopIteration < Size(data)StartYesEndFoot on Ground?WhileNoFoot on Ground?WhileStep NoIf?YesWhileInputOutputFoot Off Ground?StepNoYesOutputData Hz1/XInverse.125*XProductMin Step Size Met?Min Step Requirement NoShift IterationMin Step Size Met?NoYesYesNoOutputJust walk through it31ResultsThree ExperimentsWhat are we looking to see in each scenarioWhat does each experiment tell us?

Manual Characterization of Curves PlausibilityProduced unique data, SD

Should I include the SD charts or can I just say they are unique33Characterization of Curves Piezo Electric PlateAutomated modelingErroneous Gait speedsProduced unique coefficients

Not as big a deal as I thought they would be honestly34

Characterization of Curves Second IR SensorAutomated modelingBetter Gait Speed ModelingProper Time ShiftingHighly unique Model sets

Pattern RecognitionExperiment UsedLeast Min Squares60%Success RateCorrect Prediction/Number of steps Level WalkingDown StairsUp StairsDown RampUp RampCorrect (%)94.990.080.097.582.8# of Steps9850508070Level Walking Predictions

Level Walking Incorrectly Interrupted as Ramp Up and Ramp Down

Data Correctly Predicted as Level WalkingDown Stairs Predictions

Data Correctly Predicted as Walking Down Stairs

Down Stairs Incorrectly Interrupted as Ramp DownUp Stairs Prediction

Data Correctly Predicted as Walking Up Stairs

Up Stairs Incorrectly Interrupted as Ramp Down and Ramp Down

Data Correctly Predicted as Walking Ramp DownRamp Down Prediction

Ramp Down Incorrectly Interrupted as Ramp UpRamp Up Predictions

Data Correctly Predicted as Walking Ramp Up

Up Ramp Incorrectly Interrupted as Ramp Down and Ramp DownConclusionsCharacterization of CurvesReliableUniquePredicting Ground Types80%Not done yetFuture WorkProducing a truly portable version of the systemIntegrating system into an AFOAdding functionality of tracking long term change of patients gait characteristics for clinicians.Showing a client their progress over time!

AcknowledgmentsFunding was supplied by the RGHS RIT alliance Seed fundRochester General HospitalRichard L Barbano, MD, Ph.D., FAANAdvisorElizabeth A. DeBartolo, Ph.D.Thesis Committee Mario Gomes, Ph.D.Kathleen Lamkin-Kennard, Ph.D.Nazareth College Physical Therapy ClinicJ.J. Mowder-Tinney PT, PhD, NCSRochester Orthopedic LabsShawn Biehler, CPO

Questions?

Proposed DesignAttaches to the back of existing AFOLinear actuator shifts carriage to either side Carriage holds two individually adjustable backstops The actuator doesnt have to support the footInfrared range finderDetect terrain

Include detailed schematic only of final proposed designAccelerometersRequire a lot of data analysisDrift in integration accuracy Measuring many different thingsMost of which I am uninterested inMost of which is very noise

Same, or remove if youre not going this routeMulti Surface Sensing Ankle Foot OrthoticFunctional Block DiagramPowerGround ProfileGround Identification SensorMicro-ControllerPositionActuatorRange of MotionIfStairs or RampLevel GroundInputs out putsFit the 3 slids you are missing into this oneTalk about separating out these different criteria into ways to generate more designs48Distance Detected Level Ground vs. Descending Stairs

Feasibility talk about using gait data from other studies to generate simulations of what a distance sensor would need to seeInfra red range finders

Detectable differences

49Carbon Fiber BraceBreaks the project up into 2 areasBulk reduction in weight and size will help get patients excited about their braceSpring properties of carbon fiber

Figure 7. Carbon FiberFocus on the mechanism for this talk eliminate or keep as backupIR Range FinderRange Finder Selected GP2Y0A02Yk Image Source: http://www.technologicalarts.com/myfiles/data/gp2d120.pdf

1.5 4 8 12 24 31.5 40 59.5 216.5MaximumRangeMinimumRange

Walking Up Stairs Raw Distance Sensor VoltageReplace technical drawing with chart from next page showing rangeBody Segment DataLink No.Corresponds toLink Mass mi (kg)Link Length Li (m)Distance to COM ai (m)1Stationary Calf3.370.4730.1862Stationary Thigh7.020.4410.1883Swinging Thigh7.020.4410.2534Swinging Calf3.370.4730.2875Swinging Foot1.670.150.076