Online Event Recognition from Moving Vehicles E Tsilionis 1 , N - PowerPoint PPT Presentation

Online Event Recognition from Moving Vehicles E Tsilionis 1 , N Koutroumanis 2 , P Nikitopoulos 2 , C Doulkeridis 2 and A Artikis 2,1 1 NCSR “Demokritos”, Greece 2 University of Piraeus, Greece http://cer.iit.demokritos.gr http://www.datastories.org

Structure ◮ A two component online fleet management system. ◮ Effective integration of spatial reasoning with temporal reasoning for online CER. ◮ Evaluation on large, real-world, heterogeneous data. 1/19

System Architecture 2/19

Enrichment of Mobility Data with Weather and Points-of-interest 3/19

Weather Enrichment — How does it work? 4/19

Points-of-interest Enrichment ◮ Practically a stream-based, parallel distance join ◮ Based on spatial partitioning 2D space (in cells) ◮ Cells assigned to worker nodes (W 1 , W 2 , ) ◮ Streaming data set: assigned to enclosed cell ◮ Static data set: assigned to enclosed cell and replicated to cells at distance ≤ θ 5/19

Composite Event Recognition Overview Input ◮ Recognition ◮ Output � . . . . . . . . . . . . Event Simple Events Recognition Composite Events System . . . . . . . . . . . . happensAt ( moving ( V 1 ) , T 1 ) happensAt ( iceOnRoad ( V 1 ) , T 1 ) happensAt ( abruptAcceleration ( V 1 ) , T 5 ) holdsFor ( dangerousDriving ( V 1 ) = true , [ T 1 , T 10 )) happensAt ( stopped ( V 1 ) , T 10 ) . . . Composite Event Definitions initiatedAt ( dangerousDriving ( V ) = true , T ) ← terminatedAt ( dangerousDriving ( V ) = true , T ) ← happensAt ( iceOnRoad ( V ) , T ) , happensAt ( stopped ( V ) , T ) . holdsAt ( highSpeed ( V ) = true , T ) . 6/19

Event Calculus ◮ A logic programming language for representing and reasoning about events and their effects. ◮ Key components: ◮ event (typically instantaneous). ◮ fluent: a property that may have different values at different points in time. ◮ Built-in representation of inertia: ◮ F = V holds at a particular time-point if F = V has been initiated by an event at some earlier time-point, and not terminated by another event in the meantime. 7/19

Run-Time Event Calculus (RTEC) Predicate Meaning happensAt( E , T ) Event E occurs at time T initiatedAt( F = V , T ) At time T a period of time for which F = V is initiated terminatedAt( F = V , T ) At time T a period of time for which F = V is terminated holdsFor( F = V , I ) I is the list of the maximal intervals for which F = V holds continuously holdsAt( F = V , T ) The value of fluent F is V at time T Window ω time q 136 q 137 q 138 q 139 8/19

Input — Output events Events Description moving ( V , S ) Vehicle V is moving with a speed S stopped ( V ) Vehicle V is not moving abruptAcceleration ( V ) Vehicle V accelerates abruptly Input abruptDeceleration ( V ) Vehicle V decelerates abruptly abruptCornering ( V ) Vehicle V turns abruptly fuelLevel ( V , L ) The level of fuel in tank of vehicle V is L iceOnRoad ( V ) Vehicle V is moving in an icy road closeToGas ( V ) Vehicle V is near a gas station highSpeed ( V ) Vehicle V exceeds the user-specified speed limit Output dangerousDriving ( V ) Vehicle V is potentially moving in a dangerous way reFuelOpportunity ( V ) There is refueling opportunity for vehicle V 9/19

Visualisation of a dangerousDriving activity 10/19

Visualisation of a reFuelOpportunity activity 11/19

Experimental Setup ◮ 4 M GPS traces of moving commercial vehicles (Vodafone Innovus), 1 month ◮ Data enrichment: ◮ Weather information (frozen precipitation surface), 120 GRIB files, ≈ 7 . 4GB ◮ Points of interest (gas stations), 140K 12/19

Experimental Results Data enrichment component POI Enrichment POI Enrichment Throughput (msg/sec) Weather Enrichment Execution time (sec) 15000 800 12000 600 400 9000 200 0 6000 2 4 8 300 600 900 1200 CPU Cores Distance θ (meters) 13/19

Experimental Results CER component Avg Recognition Time (ms) Throughput (Thousand events / second) 200 1 core 250 1 core 2 cores 180 2 cores 4 cores 8 cores 200 4 cores 160 8 cores 140 150 120 100 100 50 80 60 0 1 2 3 4 5 6 7 8 40 1 2 3 4 5 6 7 8 Window size (hours) Window size (hours) 14/19

Incremental Reasoning ◮ Handle delayed arrival of input events ◮ Use of overlapping temporal windows ◮ Avoid re-computation from scratch, i.e. RTEC ◮ Update only the CEs affected ◮ Improve the computational performance 15/19

Incremental Reasoning Rule Transformation initiatedAt ( dangerousDriving ( V ) = true , T ) ← � ins � happensAt ( abruptAcceleration ( V ) , T ) , � Q i . � holdsAt ( highSpeed ( V ) = true , T ) initiatedAt ( dangerousDriving ( V ) = true , T ) ← happensAt ( abruptAcceleration ( V ) , T ) , �→ holdsAt ( highSpeed ( V ) = true , T ) . initiatedAt ( dangerousDriving ( V ) = true , T ) ← � Q i \ ins � happensAt ( abruptAcceleration ( V ) , T ) , � ins � holdsAt ( highSpeed ( V ) = true , T ) . 16/19

Incremental Reasoning Computation initiatedAt( dangerousDriving(V)=true,T ) ← ω [ happensAt( abruptAcceleration(V),T ) ] ins , [ holdsAt( highSpeed(V)=true,T ) ] Q i . time q i - ω q i-1 q i 17/19

Incremental Reasoning Computation initiatedAt( dangerousDriving(V)=true,T ) ← ω [ happensAt( abruptAcceleration(V),T ) ] ins , [ holdsAt( highSpeed(V)=true,T ) ] Q i . time q i - ω q i-1 q i 17/19

Incremental Reasoning Computation initiatedAt( dangerousDriving(V)=true,T ) ← ω [ happensAt( abruptAcceleration(V),T ) ] ins , [ holdsAt( highSpeed(V)=true,T ) ] Q i . time q i - ω q i-1 q i 17/19

Incremental Reasoning Computation initiatedAt( dangerousDriving(V)=true,T ) ← ω [ happensAt( abruptAcceleration(V),T ) ] ins , [ holdsAt( highSpeed(V)=true,T ) ] Q i . time q i - ω q i-1 q i 17/19

Experimental Results Evaluation Avg Recognition Time (ms) Avg Recognition Time (ms) 400 400 RTEC RTEC 350 350 Incremental RTEC Incremental RTEC 300 300 250 250 200 200 150 150 100 100 50 50 0 0 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 Window size (hours) Window size (hours) (a) 5% (b) 10% Avg Recognition Time (ms) 400 RTEC 350 Incremental RTEC 300 250 200 150 100 50 0 1 2 3 4 5 6 7 8 Window size (hours) (c) 20% 18/19

Summary ◮ Presented a stream reasoning online fleet management system. ◮ Evaluation illustrates the scalability of the system and its capacity to operate in real-time. ◮ Future work: ◮ User evaluation. ◮ Use of deadlines mechanism to deal with longer CE intervals (RTEC2). 19/19