Dynamic routing versus static routing Prof. drs. Dr. Leon Rothkrantz .

Dynamic routing versus static Dynamic routing versus static routingrouting

Prof. drs. Dr. Leon Rothkrantz

http://www.mmi.tudelft.nl

http://www.kbs.twi.tudelft.nl

Outline presentationOutline presentation

• Problem definition• Static routing Dijkstra shortest path algorithm• Dynamic traffic data (historical data, real time data)• Dynamic routing using 3D-Dijkstra algorithm• Travel speed prediction using ANN• Personal intelligent traveling assistant (PITA)• PITA in cars and in trains

IntroductionIntroduction

Problem definition• Find the shortest/fastest route from A to B

using dynamic route information.• Research if dynamic routing results in shorter

traveling time compared to shortest path• Is it possible to route a traveler on his route in

dynamically changing environments ?

(Non-) congested road(Non-) congested road

TrafficTraffic

Testbed: graph of highwaysTestbed: graph of highways

MONICA networkMONICA networkMany sensors/wires along the road to Many sensors/wires along the road to

measure the speed of the carsmeasure the speed of the cars

Smart RoadSmart Road

Many sensors (smart sensors) along a road Sensor devices set up a wireless ad-hoc network Sensor in the car is able to communicate with the road Congestion, icy roads can be detected by the sensors

and communicated along the network, to inform drivers remote in place and time

GPS, GSM can be included in the sensornetworks Wireless communication by wired

lamppost/streetlights

Real speed on a road segment Real speed on a road segment during peak hourduring peak hour

3 dimensional graph3 dimensional graphUse 3D DijkstraUse 3D Dijkstra

Why not search in this 3 dim. Why not search in this 3 dim. graph ?graph ?

This will become a giant graph:

- constructing such a 3 dimensional graph (estimating travel times) would take too

much time

- performance of shortest path algorithm for such a graph will be very poor

Shortest path via dynamic routingShortest path via dynamic routing

Expert systemExpert systemBased on knowledege/experience of daily cardriverBased on knowledege/experience of daily cardriver

(entrance kleinpolderplein ypenburg)

(route ypenburg prins_claus)(file prins_claus badhoevedorp)(route badhoevedorp nieuwe_meer)(exit nieuwe_meer coenplein)

Translate routes to trajectories between junctions and assign labels entrance, route, file and exit to each trajectory

Design (1)Design (1)

Schematic overview of a P+R route.

Design (2)Design (2)

Static car and public transport Static car and public transport routesroutes

Dynamic car routeDynamic car route

P+R routeP+R route

Expert systemExpert system

(entrance kleinpolderplein ypenburg)

(route ypenburg prins_claus)

(file prins_claus badhoevedorp)

(route badhoevedorp nieuwe_meer)

(exit nieuwe_meer coenplein)

Translate routes to trajectories between junctions and assign labels entrance, route, file and exit to each trajectory

Example alternative routesExample alternative routesusing expert knowledgeusing expert knowledge

Implementation in CLIPSImplementation in CLIPS

Results of dynamic routingResults of dynamic routing

Based on historical traffic speed data dynamic routing is able to save approximately 15% of travel time

During special incidents (accidents, road work,…) savings in travel time increases

During peak hours savings decreases

User preferencesUser preferences

Shortest travel timePreference routing via highways, secondary

roads minimizedPreferred routing (not) via toll routesFastest route or shortest routeRoute with minimal of traffic jams

TrafficTraffic

Current systems developed at TUDelft

• Prediction of travel time using ANN (trained on historical data)

• Model of speed as function of time average over road segments/trajectories

• Static routing using Dijkstra algorithm• Dynamic routing using 3D Dijkstra• Dynamic routing using Ant Based Control algorithm• Personal Traveling Assistant online end of 2008

NN ClassifiersNN Classifiers

Feed-Forward BP Network– single-frame input– two hidden layers– logistic output function in

hidden and output layers– full connections between layers– single output neuron


Time Delayed Neural Network– multiple frames input– coupled weights in first hidden layer for time-

dependency learning– logistic output

function in hidden and output layers

(continued)


Jordan RecursiveNeural Network– single frame input– one hidden layer– logistic output function

in hidden and output layer– context neuron for time-dependency learning

(continued)

Factors which have impact on the Factors which have impact on the speedspeed

Factors• Time• Day of the week• Month• Weather• Special events

Impact on speedImpact on speed

Time


Day of the week


Day of the week


Month


Month


Weather


Special events

Model 1Model 1

Is it possible to predict average speed on a special location and time?

Model 1Model 1

P r e d i c t o r

o x ( t )

t

d ( t )

d a ( t )

w x ( t )

p e ( t )

s e ( t )

e e ( t )

s i e ( t )

h ( t )

Model 2Model 2

Is it possible to predict average time 25 minutes ahead on a special location with an error of less then 10% ?

Model 2Model 2

P r e d i c t o r

t

d ( t )

d a ( t )

w x ( t )

p e ( t )

s e ( t )

e e ( t )

s i e ( t )

h ( t )

o x ( t - t ) … o x ( t - 2 t ) o x ( t - d t )

o x ( t ) … o x ( t + t ) o x ( t + k t )

Model 3Model 3

…

Predictor

t

d(t)

da(t)

wx(t)

pe(t)

se(t)

ee(t)

sie(t)

h(t)

ox (t) … ox (t + t) ox (t + kt)

ox-ix (t - t) … ox-ix (t - 2t) ox-ix (t - dt)

ox-x (t - t) … ox-x (t - 2t) ox-x (t - dt)

ox(t - t) … ox (t - 2t) ox (t - dt)

Test results Model 1Test results Model 1

• 6 networks tested• Tuesday• A12 in the direction of Gouda• Best results with 5 neurons in hidden layer



• 9 networks tested• Tuesday• A12 in the direction of Gouda• Best results with 9 neurons in the hidden layer


Test resultsTest results

Test resultsTest results

Results of the best performing network:

• 76% of the values with difference of 10% or less

• Average error is more than 20%• Deleting outliers: average error less than 9%

ConclusionsConclusions

• Existing research• Formula of Fletcher and Goss• Impact• Results

Current systemCurrent system

• Model (based on historical data)• Accidents and work on the road• Travel time (based on Recurrent neural

networks)• Data collection (average speed per segment, per

road)

Ant Based Control Ant Based Control Algorithm (ABC)Algorithm (ABC)

Is inspired from the behavior of the real antsIs inspired from the behavior of the real ants

Was designed for routing the data in packet switch networksWas designed for routing the data in packet switch networks

Can be applied to any routing problem which assumes dynamic Can be applied to any routing problem which assumes dynamic data like:data like:

Routing in mobile Ad-Hoc networks Routing in mobile Ad-Hoc networks Dynamic routing of traffic in a cityDynamic routing of traffic in a city Evacuation from a dangerous area ( the routing is done to multiple Evacuation from a dangerous area ( the routing is done to multiple destinations )destinations )

Natural ants find the Natural ants find the shortest routeshortest route

Choosing randomlyChoosing randomly

Laying pheromoneLaying pheromone

Biased choosingBiased choosing

3 reasons for3 reasons for choosing the shortest choosing the shortest

pathpathEarlier pheromone (trail completed

earlier)More pheromone (higher ant density)Younger pheromone (less diffusion)

AApppplication of ant lication of ant behaviourbehaviour in network in network

managementmanagement

Mobile agentsProbability tablesDifferent pheromone for every destination

Traffic mTraffic mooddel el inin oneone node node

i j k

1 pi1 pj1 pk1

2 pi2 pj2 pk2

.. ..

N piN pjN PkN

Routing tableRouting table

Local TrafficLocal Traffic StatisticsStatistics

NetworkNetworknodenode

des

tin

atio

ns

des

tin

atio

ns

neighboursneighbours

1 2 .... N

μ1;σ1; W1 μ2; σ2; W2 … μN; σN; WN

Routing tableRouting table

To forward the packets, each node has a routing table

6 8 101 0.4 0.5 0.1

2 0.7 0.2 0.1

…11 0.4 0.1 0.5

All possible destinations

Neighbours

1

4

2

3

7

9

8

10

11

65

Generating virtual ants Generating virtual ants (agents)(agents)

1. ants are launched on regular intervals

- it goes from source to a randomly chosen destination

1

4

2

1 3

7

9

8

10

11

65

11

Chosing the next nodeChosing the next node

2

1 5

2. Ant chooses its next node according to a probabilistic rule:

-probabilities in routing table;

-traffic level in the node;

2 5

11 0.4 0.6

neighbours

destination

2

Sniffing the networkSniffing the networkAnt moves towards its destination

…and it memories its path

2

11 t5

10 t4

9 t3

3 t2

2 t1

1 t0

11

8

4

3

7

9

10

11

65

11

3 9

10

2

8

The backward antThe backward ant

Ant goes back using the same path

11 t5

10 t4

9 t3

3 t2

2 t1

1 t0

1

10

1

11

10

2

4

3

7

9

65

3 9

Updating the Updating the probability tablesprobability tables

On its way to the source, ant updates

routing tables in all nodestable in 1 before update

table after update 2 511 0.4 0.6

2 511 0.8 0.2

2

8

1

10

1

11

10

2

4

3

7

9

65

3 9

SimpleSimple formula formulaee

Calculate reinforcement:

Update probabilities:

Complex formulaeComplex formulae

P’jd=Pjd + r(1-Pjd)

P’nd=Pnd - rPnd , n<>j

Map representation for

simulation

Simulation Simulation environmentenvironment

ResultsResults

Number of timesteps8,0006,0004,0002,0000

Ave

rage

sm

art r

oute

tim

e

160

140

120

100

80

60

40

20

0

Number of timesteps8,0006,0004,0002,0000

Ave

rage

sta

ndar

d ro

ute

time

180

160

140

120

100

80

60

40

20

0

Average trip time for the cars using the routing system

Average trip time for the cars that not use the routing system

Simulation environmentSimulation environment

Architecture

GPS-satellite

Vehicle

Routing system

Simulation

GPS-satellite

Vehicle

Routing system

• Position determination

• Routing

• Dynamic data

Communication flowCommunication flow

Routing systemRouting system

Routing system

Route finding system

MemoryTimetable updating system

Dynamic data

Routing

1 2 4 5 …

1 - 12 15 - …

2 11 - - 18 …

4 14 - - 13 …

5 - 18 14 - …

… … … … … …

13

2

4 5

6

7

Routing system (2)Routing system (2)Timetable

ExperimentExperiment

Personal intelligent Personal intelligent travel assistanttravel assistant

PITA is multimodal, speech, touch, text, picture,GPS,GPRSPITA is able to find shortest route in time using dynamic traffic

dataPITA is able to launch robust agents finding information on

different sites (imitating HCI)PITA computes shortest route using AI techniques (expertsystems,

case based reasoning, ant based routing alg, adaptive Dijkstra alg.)

PDAPDA

Digital AssistantDigital Assistant

Digital assistant has characteristics of a human operatorAmbient IntelligentContext awarenessAdaptive to personal characteristicsIndependent, problem solverComputational, transparent solutionsMultimodal input/output

Schematic overview of Schematic overview of the PITA componentsthe PITA components

Overview of Overview of communicationcommunication

Wireless network Wireless network layers:layers:

human human communication communication layerlayer

virtual virtual communicationcommunication

virtual virtual coordinating agentcoordinating agent

Actors, Agents and Actors, Agents and ServicesServices

Layers of Layers of communication:communication:

overlapping overlapping clouds of actors clouds of actors ( human sensors, ( human sensors, perception perception devices)devices)

corresponding corresponding clouds of clouds of representative representative agentsagents

clouds of clouds of servicesservices

Mobile Ad-Hoc Mobile Ad-Hoc NetworkNetwork

PITA system in a trainPITA system in a train

Travelers in train have device able to set up a wireless network in the train or to communicate via e-mail, connected to GPS

Position of traveler corresponds to position of trains

(de-)Centralized systems knows the position of train at every time and is able to reroute and inform travelers in dynamically changing environments

A technical view of the PITA system

The personal agent

The handheld interface model

The handheld application model

A handheld can be connected to the rest of the system by only an ad-hoc wireless connection

Sequence diagram of the addition of a new delay

The distributed agent platform architecture

User profiles

THE MAPPING BETWEEN THE USER PROFILES AND THE SEARCHPARAMETERS

The route plan to Groningen Noord

Search times

Dynamic routing versus static routing Prof. drs. Dr. Leon Rothkrantz .

Documents

dynamic routing results

dynamic route information

d dijkstradynamic routing

real time datadynamic

shortestfastest route

time performance

shorter traveling time

toll routesfastest route