Always-on Considerations for Mobile Computing · PDF fileAlways-on Considerations for Mobile Computing ... Intel Labs Desktop grows across 4 -displays ... Schematic Block Diagram Compass

1

Intel Labs

Always-on Considerationsfor Mobile Computing

DAC’10 Workshop

Mobile & Cloud Computing

14th June 2010

Roy Want

2

Intel Labs

Overview

• Motivation for Mobile Sensing & Cloud Services

• Requirements for Sensing• Always-on behavior• Power Considerations• Proposed Architectural Solution• Conclusion

SensingWhere I am,

and

What am I doing?

3

Intel Labs

iPhone Sensors:Accelerometer, GPS, Compass, Gyro, Light, Touch Screen

iPod made sensing mainstream for mobiles

4

Intel Labs

Many Projects are exploring the use of Sensors: Common Sense (Intel) & Sensor Planet (Nokia)

5

Intel Labs

Cloud Services & Context

Twitter LooptFacebook

6

Intel Labs

Augmented Reality Services

1. Vendor list applicationaugmented using “MonicleMode”

2. GPS finds location

3. Compass tells the devicethe direction viewed

4. Augmented reality based apppresents composite image

7

Intel Labs

Local Clouds: Context-Aware Discovery

Lime Sharing its display

with Plum &Orange

red

Red Shares itsAccelerometerSensor with

Orange

Computers

Resourcethat canBe shared

Interface Shows Spatial Arrangement

JOIN-THE-DOTS TO SHARE

Shows Direction & Movement of Devices

8

Intel Labs

Local Clouds: Multi-display Composition (a)

Ultrasonic Ranging: Ping Listen-RX (Direction, Range)Accuracy = better than 1cm over 1m

9

Intel Labs

Desktop grows across 4-displays – Shown at CES 2009

Local Clouds: Multi-display Composition (b)

10

Intel Labs

Requirements for Mobile Sensing • A single serialized connection• Rationalized sensing interface (uniform syntax)• Self describing sensor interface (catalog of sensors)• Accurate time-stamped sample data• One-shot “get” sensor reading or continuous “streaming”• Power-aware operation• Extensible multi-sensor support

– Independent periods per sensor

• Multi-application support– Independent streams per sensor (supports multitasking)

• Buffering (N-samples burst transfer)• Threshold set-points (e.g. events only above/below threshold)• Conditionals (e.g. trigger one sensor reading based on another) • Operators on streams (e.g. low-pass filter)• Error reporting (e.g. cannot meet QOS requested)

11

Intel Labs

Low-power operation

TIMER

(S-Time)

QueueHead

Sensor: ACC

Period: w

S-time: a

SID: 1

Sensor: Gyro

Period: x

S-time: b

SID: 2

Sensor: ACC

Period: y

S-time: c

SID: 3

On interrupt: Remove head item; Sample & TX value: AddQueue; New S-Time=NextSensor(S-time);

Sleep (wake on int)

Sensor: light

Period: z

S-time: d

SID: 4

AddQueue S-Time= RealTime + (SID) Period t; Insert in Order List

DelQueue

Ordered List Based on Sample Time (S-Time)

Remove SID from List

SLEEP

12

Intel Labs

Sensor Hub Co-Processor Model

Provides Sensing (and hence context) Co-processing Capability

Sensor Actuator API

(SAAPI)

HOST Computer

Sensor/Actuatorco-processor

Interface

13

Intel Labs

Sensor-Bus Architecture

Co-Processor

ACC TEMP GYRO GPS

Sensor/ActAPI

COMMUNICATION BUS (e.g. COM, DBus)

APPLICATION #1 APPLICATION #2 APPLICATION #N

Host Processor Serial

Sensor-Hub

MAGBAR

Time Stamp Filter/DSPBurst Mode Threshold

I2C, SPI, UART

ERR Report

Stream/Poll Power

14

Intel Labs

Sensor Hub or ContextCardCC3.0 Feature Specification

Processor & SupportOn-board processing

– low-power ATMEL microprocessor[~ ARC capabilities]

– JTAG programming pads– Silicon ID– 3-color LED– Switches 1) Reset 2) State

Interfaces & Radios– USB2.0 interface– Near Field Communication (NFC)

interface

Sensors– 3-axis Accelerometer x2– 3-axis Compass/2D Magnetometer– 3-axis Gyroscope– Barometer– Thermal sensor– 12-input touch panel interface– Wii IR detector– IrDa interface x 4 (each face)*– Ultrasonic Transceiver x 4*

Actuators– Vibration

compatible with Immersion

* For Relative Ranging/Location Determination

15

Intel Labs15

Context Card 3.0: Schematic Block Diagram

CompassHMC6352

NFC(PN531)

RGBLED

RESETSW

STATESW

JTAGPADS

USB2.0

SLAVE( UX50)

MicrocontrollerUSB

Serial

NFCCoil

8-ElementTouch Sensor

2D Gyros

IR Transceiver x 4

MatchingCircuit

Multiplexer.

802.15.4 Radio

Serial

I2C

SPI

3-AXIS Accel.(x2)

UltrasonicTX/RX x3SRAM

Uart1

UART2

UART3

GPIO

SPI0(0)

I2C BUSSPI1(0)

ADC

SPI1(1)

BarometerSMD500

Expansion Con21P & Con31P

ThermalSensor

SPI1(2)PWM

LightSensor

Actuator & Vibrator

ADCSensor Bus

Silicon-IDDS2411

TOHOST

WiiSensor

iPhone Interface

16

Intel Labs

A Prototype Mobile Sensor-Hub CC3.0

3.5” 4.5”

17

Intel Labs

Power Measurement for each sensor

ADC

Multiplexer

VCC

Sensor

Reference load

1

N

10R – 0.1R – 0.01R

20 milli-ohm

VCC

out

I2C to Processor

I2C to Processor

18

Intel Labs

Using CC with a PC via Hypertern

<device name>“id” = < number >“op” = < “get” | “set” | “stream” >“val” = < “on” | “off”>“samples” = number “period” = < mS >“power” = < “on” | “off”>“threshold” = <“ <hi>:<low>” >“mode” = < “delta” | “abs” | “up” | “down” >

e.g. acc id=1 op=stream val=on samples=10 period=1000

USBConsoleHyperterm

19

Intel Labs

Rationalized Sensor-Hub API

EVT:098789 acc val=120:50:-10 p=22 s=1

CMD: acc op=stream val=on period=500 power=on threshold=100:100 cid=102

REP: acc op=stream val=on cid=102

Command String

NOTE: Period in mS, Power in mW

Reply String

Event String

EXAMPLE: ACCELEROMETER API

Text-based, key/value pairs

20

Intel Labs

Context Card Serial Command Examples• COMMAND: led id=2 op=set val=on

RESULTS: RES:systime led id=2 op=set val=on (setting green led on)

• COMMAND: act id=1 op=getRESULTS: RES:systime act id=1 op=get val=off (getting value of actuator)

• COMMAND: therm id=1 op=get val=on samples=1RESULTS: RES:systime therm id=1 op=get samples=1 val=28.00(getting temperature = 28 C)

• COMMAND: acc id=1 op=get val=on samples=1RESULTS: RES:systime acc id=1 op=get samples=1 val=9:-3:131(getting accelerometer reading for X, Y, Z)

21

Intel Labs

Sensor HubControl Panel

Python implementation for Windows & Linux

22

Intel Labs

Future mobile architectures will benefit from:

• A coprocessor based sensor-hub subsystem

• Always-on operation

• Industry has an opportunity to create a standard sensor-hub reference design that can be reused across many mobile platforms

Cloud services will be enriched by, and benefit from, ubiquitous mobile sensing

Conclusion & take-away messages

23

Intel Labs

Thank you!

Q & A

24

Intel Labs