Notes from Mik Lamming's talk

Here's the notes of the talk today with Mik Lamming that I wrote with my laptop.

Sensing Behaviour
Mik Lamming, HP Labs

Before Mik began to talk, he discussed about how before he got into research, he used to work in Photoshop. He met Roy Want (who now works at Intel Research) from Olivetti with the Active Badge. Active Badge is a device which looks like a badge and is equipped with a radio for sensing where you are in the office environment. It was one of the first ubicomp systems deployed. Mik then talked about his work on the Forget-me-not which he worked on at Xerox PARC.

Forget-me-not
Xerox, 1993

- Pervasive over space and time
- Proximity information better than location
- Showed BirdDoc to Xerox, person under surveillance
- Roy Want worked in Xerox to help Mik with Active Badge with a display -> PARCTab was invented
- Log people’s behaviour, who you were with, where you are, log people’s activities by wearing these small devices – they embedded these devices everywhere in the lab
- Built system called Forget-me-not as a memory prosthesis unit
- Want to remember things in the home not just in the lab
- Xerox said no possible use for this system
- Repurposed the system as document retrieval system, how to retrieve documents
- Lessons learned:
- people are working on systems with absolute position (have to make uplink to wireless network, that burns power, it’s more expensive to do that than peer-to-peer pinging, that was their theory)
- there was lack of sympathy to tracking people’s behaviour eg. Participant would come back and the device didn’t work because it was wet
- invasive and privacy – dumb and evil
- installing this infrastructure was extremely tedious
- lots of these applications need to work in your own house, single user value proposition, rather than lots of users

Alzheimer’s and Caregivers Work

This was Mik's work that he talked about while working at HP Labs.

- did caregiver interviews in collaboration with Prof. Linda Nichols, UT Memphis
- caregivers are typically the elderly, they have caregiver burnout
- expensive cost of caregiving
- analyzed logs of data
- Alzheimer’s patient behaviour can fluctuate from day to day
- Designed a system for sensing
- Zone detection eg. In bathroom, in bed,near stove, fire
- Proximity
- Constructed a design rationale for environment sensing for monitoring environment

Out of this came the Minder concept.

Minder Concept
- Personal technology for sensing personal and social context while providing control and privacy of information
- Embed them in everyday things that you carry, wear
- P2P system
- Always on and attentive, yet long-lived, a year without battery change, wear it day and night
- Backed by 3rd party brokered utility services
- Proximity matters without central system to be better than Forget-me-not
- Opportunistic interactions, see proximity and upload the log to a portal for timeline analysis
- Pattern recognition language to apply to the minder device
- Make a network that can’t be heard by others based on shared authentication (privacy model), next version will use UWB

At this point, Mik passed around the prototype being deployed on caregivers and Alzheimer’s patients, very small device with a chip and using IR called the SPEC.

Using proximity for monitoring detection of Alzheimer’s patient
- address problems like leaving the stove on
- attach these devices to stove knob, and find out that it has been moved
- how to build system that will keep going long enough, and easy to change battery so don’t lose tracking information

Sensing own behaviour on a typical day
- Mik showed graph of what he did, he deployed SPECs in different locations
- From here, can tell what Mik did everyday (probably more than he wanted to know and what the whole world to know)
- Wrote analysis software to take the interaction graph to translate into human interpretable terms
- Why keep diary of behaviour? eg. For drug trials
- Does this reveal too much private information, figure out too much about yourself and share with the world?
- By talking to other devices in the environment, keep log on your device

Recognizing and responding to situations in real time
- tested this with a kid
- event queries may be downloaded
- downloaded a pattern
- deployed the SPECs on backpack, scooter, garage, desk and bedroom of the kid's belongings
- could detect whether he forgot his scooter by having LED emit on the SPEC

What learnt from Minder s1
- can do lot of useful things with proximity sensing, and simple pattern recognition
- power, adding sensors saves power if gives hints when device can sleep
- by using knowledge of behaviour, then can turn off power when necessary
- form factor, can now do 24X7 tracking, need to be invisible and fashionable, not geeky
- try to embed in things that people use and wear, example embed in watch

Minder s1 / “Firefly 1”
- wristwatch form factor with RF, accelerometers/magnetometers for walking/orientation
- doing another iteration of this device with range finding to do time of flight to use chirp-mode radio and not UWB
- moving ahead with the home care
- lots of uses that people want for this, eg. Google way for providing ads based on human behaviour

Mik's sensing behaviour work with the SPEC reminds me of the research project I worked on opportunistic interactions using Palm handheld devices. In this case, we used Palm Tungsten T handhelds equipped with Bluetooth and wrote software to make pair-wise connections with other Palms and Bluetooth devices nearby. Of course, we never did achieve great battery life like the SPEC which is like in the order of couple of months. Our Palms only lasted 6 hours on a good day before having to be charged.

The wireless radios out there like Bluetooth, UWB, Wi-Fi were not really designed for opportunistic spontaneous networking. They are power-hungry and take a long time to establish connection. A custom radio needs to be designed for something like this that has almost zero-power.