The CarTel Project Lewis Girod M.I.T. Computer Science & Artificial Intelligence Lab cartel.csail.mit.edu
MIT/CSAIL MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) Entrepreneurial approach Spawned over 100 technology companies Industrial Liaison Program develops relationships with industry Interdisciplinary approach Primarily within Electrical Engineering and Computer Science Dept Collaboration and overlap with Mathematics, Cog Sci, Mechanical Engineering, Civil and Environmental Eng, Media Arts and Sciences, Health Sciences, Bioinformatics
CarTel Project Active project since 2002 Funding from NSF CNS-0205445, CNS-0520032, CAREER-0448124, NSF CPS, Quanta Computer, and Google Principal investigators Hari Balakrishnan Sam Madden Project Theme: Transportation-related applications Planning, traffic mitigation, end-user applications Leverage disruptive advances in mobile networks and embedded computing
Many Traffic Data Sources Installed loop sensors GPS sensors in service vehicles Transport service vehicles (taxis, busses, etc) Commercial fleets (delivery networks) Government vehicles Personal mobile phones Smartphones with GPS Cellular location services Each has benefits and liabilities CarTel technology applies to entire spectrum Innovation in both data collection and data analysis
CarTel Technology Projects: Collecting and Analyzing Traffic Data Instrumentation projects Taxis: Boston limo service; Singapore taxi data iphone and Android applications Phone data mapped to Chicago CTA bus routes Analysis of cellular signal strength + GPS Analysis and data mining projects Singapore taxi data analysis and traffic models vtrack (WiFi access point data traffic
Application to 3 challenges Encourage participation, provide info to public icartel smartphone app and web portal Chicago CTA crowdsourced bus routes Identify urgent needs, support planning Large-scale traffic data collection and mining VTrack, CTrack, phone applications Ultra low cost probe devices Smartphones, feature phones Ultra-low-cost dedicated probes
Sam Madden & Hari Balakrishnan icartel2 iphone App
icartel Web Commute Portal Current traffic delays
VTrack: Accurate Traffic Modeling Map matching Determine best trajectory Accurate delay estimation Match observation to segment to extract delays Predictive model Highly scalable and parallel cloud implementation Maximizes accuracy per probe asset VTrack: Accurate, Energy-Aware Road Traffic Delay Estimation Using Mobile Phones. Arvind Thiagarajan, Lenin Ravindranath, Katrina LaCurts, Sivan Toledo, Jakob Eriksson, Samuel Madden, Hari Balakrishnan, in Proc. 14th ACM SenSys, Berkeley, CA, November 2009.
CTrack: Map Matching and Delay Estimation from Raw Cellular RSSI Data Raw points (placelab) HMM fingerprints to grid sequence Smooth + interpolate HMM smooth grid to map Accurate, Low-Energy grid Trajectory sequence Mapping for Mobile Devices. Arvind Thiagarajan, Lenin S. Ravindranath, Hari Balakrishnan, Samuel Madden, Lewis Girod, Proc. NSDI, Boston, MA, 2011.
Technical Approach Modeling, prediction, analysis VTrack / CTrack delay estimation Traffic-aware routing Servers in the cloud Location-based vehicular services Highly granular raw position collection Aggregation provides environmental info Details Smartphones enable personalized in vehicles service Internet
Singapore: Delay Model: AYE (4037) Distance: 186 meter Sejoon Lim & Daniela Rus
Chicago CTA Bus Route Experiment Goal: Deduce and predict actual bus route timing based on rider phone reports Match rider observations against bus routes System based on iphone application + server based data mining Made use of GPS and acceleration sensors Arvind Enables Thiagarajan, James riders Biagioni, to Tomas better Gerlich, anticipate and Jakob Eriksson. bus Cooperative transit tracking using gps-enabled smartphones. In SenSys, pages 85-98. ACM, 2010. arrival
Reducing cost of probes Smartphones/Java phones works out of the box Feature phones need provider cooperation to gather RSSI data; analysis via Ctrack Lowering costs of dedicated embedded solutions Probe data alternatives: GPS probe costs as little as $15 in high volume Cellular RSSI via passive monitoring of cellular signaling WiFi RSSI + opportunistic upload Data upload mechanisms: Opportunistic WiFi
Overview Traffic mitigation technology to Scalably process raw data from mobile phones to produce accurate traffic models Produce user-specific traffic-optimized routes and schedules Reduce overall costs of data collection and analysis Enables commute optimization for consumers and tools for planners http://cartel.csail.mit.edu