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

▶

Feb 11, 2024 326 likes •569 views

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

SLIDE 1

Online Event Recognition from Moving Vehicles

E Tsilionis1, N Koutroumanis2, P Nikitopoulos2, C Doulkeridis2 and A Artikis2,1

1NCSR “Demokritos”, Greece 2University of Piraeus, Greece

http://cer.iit.demokritos.gr http://www.datastories.org

SLIDE 2

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

SLIDE 3

System Architecture

2/19

SLIDE 4

Enrichment of Mobility Data with Weather and Points-of-interest

3/19

SLIDE 5

Weather Enrichment — How does it work?

4/19

SLIDE 6

Points-of-interest Enrichment

◮ Practically a stream-based, parallel distance join

◮ Based on spatial partitioning 2D space (in cells) ◮ Cells assigned to worker nodes (W1, W2, ) ◮ Streaming data set: assigned to enclosed cell ◮ Static data set: assigned to enclosed cell and

replicated to cells at distance ≤ θ

5/19

SLIDE 7

Composite Event Recognition Overview

Input ◮ Recognition ◮ Output

Event Recognition System Composite Event Definitions

Simple Events . . . . . . . . . . . .

happensAt(moving(V1), T1) happensAt(iceOnRoad(V1), T1) happensAt(abruptAcceleration(V1), T5) happensAt(stopped(V1), T10)

. . .

Composite Events . . . . . . . . . . . .

holdsFor(dangerousDriving(V1) = true, [T1, T10)) 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

SLIDE 8

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

SLIDE 9

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 time q136 q138 q137 Window ω q139

8/19

SLIDE 10

Input — Output events

Events Description Input moving(V , S) Vehicle V is moving with a speed S stopped(V ) Vehicle V is not moving abruptAcceleration(V ) Vehicle V accelerates abruptly 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 Output highSpeed(V ) Vehicle V exceeds the user-specified speed limit dangerousDriving(V ) Vehicle V is potentially moving in a dangerous way reFuelOpportunity(V ) There is refueling opportunity for vehicle V

9/19

SLIDE 11

Visualisation of a dangerousDriving activity

10/19

SLIDE 12

Visualisation of a reFuelOpportunity activity

11/19

SLIDE 13

Experimental Setup

◮ 4M 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

SLIDE 14

Experimental Results

Data enrichment component

200 400 600 800 2 4 8 Execution time (sec) CPU Cores POI Enrichment Weather Enrichment 6000 9000 12000 15000 300 600 900 1200 Throughput (msg/sec) Distance θ (meters) POI Enrichment

13/19

SLIDE 15

Experimental Results

CER component

50 100 150 200 250 1 2 3 4 5 6 7 8 Avg Recognition Time (ms) Window size (hours) 1 core 2 cores 4 cores 8 cores

40 60 80 100 120 140 160 180 200 1 2 3 4 5 6 7 8 Throughput (Thousand events / second) Window size (hours) 1 core 2 cores 4 cores 8 cores

14/19

SLIDE 16

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

SLIDE 17

Incremental Reasoning

Rule Transformation

initiatedAt(dangerousDriving(V ) = true, T) ← happensAt(abruptAcceleration(V ), T), holdsAt(highSpeed(V ) = true, T).

→

initiatedAt(dangerousDriving(V ) = true, T) ←

happensAt(abruptAcceleration(V ), T)

ins ,

holdsAt(highSpeed(V ) = true, T)

Qi.

initiatedAt(dangerousDriving(V ) = true, T) ←

happensAt(abruptAcceleration(V ), T)

Qi\ins ,

holdsAt(highSpeed(V ) = true, T)

50 100 150 200 250 300 350 400 1 2 3 4 5 6 7 8 Avg Recognition Time (ms) Window size (hours) RTEC Incremental RTEC

(a) 5%

50 100 150 200 250 300 350 400 1 2 3 4 5 6 7 8 Avg Recognition Time (ms) Window size (hours) RTEC Incremental RTEC

(b) 10%

50 100 150 200 250 300 350 400 1 2 3 4 5 6 7 8 Avg Recognition Time (ms) Window size (hours) RTEC Incremental RTEC

(c) 20%

18/19

SLIDE 23

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