motus | technology overview statsports may 2019 v1-0...motus server workload calculations workload...
TRANSCRIPT
B I O M E C H A N I C S W O R K L O A D P R O D U C T S
4-SENSOR GAIT ANALYSIS MODULELICENSED FOR SHOE-FITTING
FIRST WEARABLE APPROVED FORMLB IN-GAME USE (2015)
T E C H N O L O G Y D O M A I N E X P E R T I S E
HIGH-DYNAMICRANGE 6-AXIS IMU
+ FIRMWARE
FULL-BODY BIOMECHANICS PHYSICS ENGINE AND SENSOR
FUSION LIBRARY
NEURAL-NETMOVEMENT CLASS-IFICATION ENGINE
IN-HOME PHYSICAL THERAPYREMOTE KNEE WORKLOAD MONITOR
CLASS 1 MED-ICAL DEVICE
REAL-TIME MOBILE3D VIEWER AND MOVE LIBRARY
FW WL
WORKLOADCOMPUTATION
CLOUD SERVICES
Te c h n o l o g y O v e r v i e wS TAT S p o r t s | M o t u s
J u n e 2 0 1 9
QB’
s,KI
CKER
SFI
ELD
ERS,
BOW
LERS
SPIK
ES,
SERV
ES
H A R D W A R E S P E C I F I C AT I O N S
I N D U S T R Y ’ S S M A L L E S TI N E R T I A L S E N S O R
BATTERY LIFE (120 mAh Rechargeable Li-Po)
8 hours for Event Trigger Mode3 hours for Logging Mode4.5 hours for Streaming Mode6 hours Hybrid Event Trigger/Streaming Mode
••••
DATA SAMPLING MODES
Logging Mode: Programatically log 6-axis data sampled at 1000 Hz.Event Trigger Mode: Stores 5 seconds of 6-axis data sampled at 1000 Hz after event detection algorithm triggers.Streaming Mode: Sample and stream 6-axis data at 100 Hz in “real-time” with 30 ms latency
•
•
•
6-AXIS INERTIAL MEASUREMENT UNIT (IMU)3-axis Gyroscope: +/- 4000 dps3-axis Accelerometer: +/- 24 g’s
••
BLUETOOTH (BLE 4.0, 2.6 kb/s)
4 seconds per Event Trigger30 ms latency for Streaming Mode40 minutes per 2 hour Logging Session
•••
C U R R E N T F I R M W A R E A R C H I T E C T U R E S
Live Streaming Mode - High Power Consumption
Sends SamplesEvery Connection
Interval (30ms)32 kHz Precision Clock
(20 ppm drift)
BLE FLASH
ACC GYRO
RAM MCU
MCU Samples 6-axisdata plus timestamp at100 Hz, and sends 3-4samples of data to perhipheral every 30ms.
100 HzOccassional Usage
of BLE & Flash
Sensor1
Sensor2
t = 0.0 s t = 20 minutes
Accurately/time-synced sensor data received by phone after each sample.
Sens
or 1
Dat
aSe
nsor
2 D
ata
Program Trigger Mode - Moderate Power Consumption
1 kHz Clock (Large Drift)
Clock Triggers x3Every 10 minutes
BLE FLASH
ACC GYRO
RAM MCUStore 5 seconds of data from RAM toFLASH upon trigger
Event Triggert = 4.0 s
t = 0.0 s t = 5.0 s
Programatically TriggerBased on Timer orEvent Detection Algorithms
Intermittent FLASHUsage
t = 0.0 s t = 16.0 s
t = 16.8 s
Resize toknown time
t = 0.0 s t = 16.0 s
t = 16.8 s
Resize toknown time
10 Minute Gap
Phone Trigger Mode - High Power Consumption
1 kHz Clock (Large Drift)Event Trigger
ACC GYRO
BLE FLASH RAM MCUStore 5 seconds of data from RAM toFLASH upon trigger
Event Triggert = 4.0 s
t = 0.0 s t = 5.0 s
t = 0.0 st = 16.0 s
t = 16.8 s
Resize toknown time
B I O M E C H A N I C S D ATA O U T P U T
SENSOR FUSION OUTPUTSensor Linear Acceleration, Velocity, Position (XYZ)Sensor Angular Acceleration, Velocity, Orientation (XYZ, Euler Angles)
••
TRANSLATED KINEMATICSObject Velocity (i.e. ball, racket, hand)Joint Kinetics (i.e. Flexion Torque, Valgus Torque, etc.)
••
MULTI-SENSOR JOINT KINEMATICSUpper Extremity: Elbow Flexion, Shoulder Internal/Ex-ternal Rotation, Abduction, Flexion/Extension.Pelvis/Torso: Flexion/Extension, Lateral Flexion, Rota-tionHip: Flexion/Extension, Internal/External Rotation, Ab/AdductionKnee/Ankle: Flexion
•
•
•
•
KEY FRAME IDENTIFICATIONPoint of Contact (i.e. ball release, ball contact)Point of Movement Initiation (i.e. load, peak back-swing).Local/Global Maximum/Minimum Detection
••
•
W O R K L O A D C L O U D S E R V I C E S
WORKLOAD MEASURESDaily Workload (Exponential sum of Torque, Veloc-ity, etc.)Acute Workload: Rolling 7 day average of daily workloadChronic Workload: Rolling 28 day average of daily workloadAcute:chronic Ratio (ACR): Acute workload divided by chronic workload
•
•
•
•
WORKLOAD SIMULATORProject 28-days of future workloads that increase chronic workload and limit high ACR exposure.Designate pre-game warmup workload routines.Simulate game-day workload and effect on ACR.Auto-generate workload limits on game-dayConvert game-day workload limits to actionable activity limits (after warmup workloads are sub-tracted).
•
••••
C U R R E N T P H Y S I C S E N G I N E A R C H I T E C T U R E
x,y,z Acc
x,y,z Gyro
x,y,z Acc
x,y,z Gyro
SENSOR FUSION& CALIBRATION
INVERSE DYNAMICSPHYSICS ENGINE
ELBOW VALGUS TORQUEEVENT DETECTION
FULL SAMPLING
ELBOW TORQUE WORKFLOW
P R O P O S E D A R C H I T E C T U R E
ELBOW TORQUE WORKFLOW
1 Capture Baseline of Full-Sampled Data 2 Train Neural Net for Torque w/ Sensor Data
STANDARD PROGRAM TRIGGER FW VIA TENSORFLOW INTEGRATION
3 Load Player-Specific Models on New FW
Requires FW re-write to handle flash memory blocks and peak detection after event detection.Requires creation of integer based math (not float).Requires creation of BLE Characteristic of count, peak torque, and arrays for APEX compatability
•
••
NEURAL NETWORK TRAINING OF ELBOW TORQUE
PRO
OF
OF
CON
CEPT
(L
AB
VALI
DAT
ION
)
E X PA N D E D P R O O F O F C O N C E P T
FIRMWARE BIOS/MCU CONTROLLER
PRE-
LOAD
ED N
EURA
LN
ETW
ORK
WEI
GH
TS
POST
-TH
ROW
PEA
KTO
RQU
E CA
LC (N
N)
1KH
z SA
MPL
ING
& E
VEN
T ST
ORA
GE
Metadata6-axis 1 KHz
Rep Counting
Metrics (velocity, etc)
3D View Data
User IDSession Data
MOTUSSERVER
Workload Calculations
Workload Prescriptions
Advanced Analytics
MOTUS SENSOR
APEX SENSOR
FIRMWARE BIOS/MCU CONTROLLER
3-AX
ISAC
CEL.
IMU
/GPS
ON
-BO
ARD
PYSI
CS C
ALCU
TIO
NS
BLU
ETO
OTH
LE 4
.0Peak Torque & TimestampBLE Characteristics
SENSOR FUSION& CALIBRATION
INVERSE DYNAMICSPHYSICS ENGINE
Workload Display
MOTUS POST SESSIONiOS APP
BLE SMARTWATCH
Rep Counting
Workload Display
Prescription Compare
STATSPORTSWIFI BEACONS
STATSPORTSREAL TIME IPAD APP
STATSPORTSSERVER
Workload Calculations
Workload Prescriptions
Advanced Analytics
Workload Prescriptions
Neural NetWeights
DATATX/RX
DATATX/RX
User IDSession Data
Workload Prescriptions
WorkloadPrescriptions
STATSPORTS DATA
MOTUSWEB-DASH
SONRAWEB-DASH