Subject:  RESCUE MEETING

Date:       01/23/04

Loc:        URP

Author:  AM

 

 

1.   Summary

Discussion of imminent visit by sponsors.  Outline of story and demos to be built in preparation.

2.   Agenda

 

3.   Discussion

 

SM:         *Site visit March 19. “Pre” site visit.

                *By March 1: A set of demos.

                Darleen Fisher, PM, RI.

                Bhavani – her boss.

                9am-5pm all day meetings.

                *Assign handful of students to meet with them.

                Review proposal:

                UCSD is ahead of us. (schedule-wise)

                Need to match their story.

                Primary goal: Research infrastructure grant (equipment)

                [White board]

                [Inner circle: people with 100 handhelds.

                Outer circle: Videocams paired with network equipment.

                Rescue office: Service, store data.

                Sync with UCSD campus.

                Mobile cyber-shuttle.]

                Hi end vs low end cameras – pan, zoom, etc.

                Need to follow the script on this.

                What will we demo.

                CAMAS – now a few months old.

                What’s doable in 1.5 months?

                Ganz looked at our demo.

-          Integrated with voice-activated system.

Navigate problem tree with voice commands.

NV:         *Borrow voice stuff from Ganz.

SM:         *Dmitri* talk to Ganz.

NV,SM:  More UCSD-UCI coordination.

SM:         *Use people 100% on Rescue. Dmitri, Amnon.

                *Campus-level real data.

                *Facilities (Greg Fountain), NACS (Dana Rude), Police, Parking

                *AM* move this forward. Next week get people in same room, in Cal-IT.

                *Within 1 week, get at least one data set.

                Data source: bulding plans.

                *Ram* talking to them.

                Geo-register with PDAs.  Phone & cam.

                *Buy 1 or 2 PDAs.

NV:         *Get cost.

CB:          Speed of algorithms.

-          Take hours, may not get to RT.

SM:         Canned.

                Before & after.

NV:         Localize in space and time.

CB:          *CB* be in on meeting with UCI people.

All/Chen: Blue sticks/emergency…

CB:          Need to simulate emergency?

                CERT = Community Emergency Response Team.

                *Leverage them.

SM:         ? has a team of CERT people.

                Agenda for social science.

                Richard Matthews – Group is ?

NV:         Tie in Padhraic’s stuff.

Sridevi:   Trajectory models.

SM:         *Integrate into CAMAS story.

CB:          Modify to do group size detection.

SM:         Learning a model of occupancy.

CB:          Condition on occupancy.

                Tells start and end.

                Piecewise Brownian motion…

                Story more important than demo.

Sridevi:   Prerecorded data.

SM:         Trajectory analysis.

                10 x 30 second segments of video.

NV:         Tie into reliability model?

CB:          To do anomaly detection.

SM:         Some source > analysis > event extraction.

All:          Demo first vs story first.

CB:          Use augmented spaces.

                Additional to physical space.

                Off the shelf technology.

                Easy to demo.

                Other patterns, associations.

                Eg, types of system failures.

SM:         Improve the CAMAS demo.

                Add real data set.

                Make it student oriented?

                Sororities, fraternities, clubs.

CB:          UCSD did some stuff.

SM:         * [CAMAS1 = DEMO #1]

                *March 1 demo system.

                Camas was mainly an analysis demo.

                *TEAM:

                Dmitri – leader

                Amnon

                Ram

                Haimin

                Jehan (1/2 ?)

 

SM:         * [DEMO #2]

                *[PRIVACY PRESERVING DATA COLLECTION]

                (Video surveillance)

                *Livewave.com

                Video cam for bio, chem. Surveillance.

                (Multi-source validation)

                Canadian – monitor hockey games…

?:             RF ID tag for tracking.

SM:         Merge models.

Jehan:     Equipment arrives next week.

                *Need Cal IT to submit…

SM:         *Kaushik worked on a 3 VC system.

Jehan:     One open area, one room.

Iosif:       *Find cam system.

SM:         Hide identity, till someone breaks a rule.

                Story of someone fired for too many coffee breaks.

GS:          Nitesh

SM:         Cryptography.

GS:          BYU, derivatives of Trust-Builder.

                Go ask Magda?

SM:         *TEAM:

                Jehan – leader

                Mahesh

                Nitesh

                ImageCatvidproc experience?

 

SM:         *[DEMO #3]

                *[ALTERNATE 911 SERVICE]

                High call volume.

                Dynamic analysis.

                Disambiguate calls.

                Build larger model.

                Same event or not.

                Set of questions to ask.

                Yiming – working filtering.

                Would fly well with response community.

                [Event E1 > S, T, P1, P2, P3.  E2, E3 …]

                Space, Time, Property1, Property2, etc.

                Call > Compare to existing events > Extract.

                If 95% accuracy, major value added, hot stuff.

NV:         Classification problems.

                Can do some of this a priori.

SM:         Forget voice for now.

                Extract from NLP?

CB:          Examples of transcribed 911 calls.

                Some available in news articles.

                News: accident logs.

                ? websites.

SM:         Claim is that a data-driven approach is better than the complex NLP done up to now.

YM:        Event generation.

SM:         Information extraction expertise – Amnon.

                Alternate 911 is part of analysis and store [Rescue Office, in earlier diagram].

 YM:       [Argument for YM’s project, approach.]

                Assume features are already extracted.

                Automatically prioritize, rank calls.

                Given type of emergency.

                Relative priority of calls coming in.

SM:         Triage to individual = OK.

                Personalize.

                Filter.

                Why real time?

                From speech?

                Start with keyword extraction.

                Space, time, type.

NV:         Tie to Ganz notification work.

                Collection, analysis, dissemination.

                [Debate with SM]

SM:         *TEAM:

                Yiming

                Qi

                Kemal

                Chen?

                Amnon?

 

NV:         4th loop sensor data?

All:          Discussion of similarity of CAMAS vs. Alternate 911.

 

NV:         *[DEMO #4]

                *[COLLECTION DEMO]

                Sensor networks.

                Live loop sensor data.

                Show map, data coming in.

                Display data.

                Display collection.

                Cost-effective real-time policies.

 

               

Xingbo:  [Presentation: Berkeley Traffic Management Ctr Website]

                http://www.pems.eecs.berkeley.edu

                22,000 sensors deployed on all California highways.

                TMC = Traffic Mgmt Ctr.

                4 levels:

                (4) TCM

                (3) Loop detector

                (2) Vehicle station

                (1) Sensor level

                + Link communications

                Vehicle speed:

(1)     by multiple detectors

(2)     by occupancy and flow.

Trafficdodger.com

SM:         2GB/16 hours.

All:          Discussion of data demo, new angle on this traffic data.

SM:         [Presentation: Earthquake before and after slides, etc]

                Video could be used to do large-scale assessment in real time.

 

 

4.   Action Items

 

All:          Note team assignments above, get demo work going.

                Note starred items and others above.

AM:        Set up meeting with UCI facilities, etc.

Dmitri:    Talk to Ganz.

Ram?:     Buy PDAs.