This work presents an original approach to regulate traffic by using an on-line system controlled by an EA. Our proposal uses computational spots with WiFi connectivity located at traffic lights (the Red Swarm), which are used to suggest alternative individual routes to vehicles. An evolutionary algorithm is also proposed in order to find a configuration for the Red Swarm spots which reduces the travel time of the vehicles and also prevents traffic jams. We solve real scenarios in the city of Malaga (Spain), thus enriching the OpenStreetMap info by adding traffic lights, sensors, routes and vehicle flows. The result is then imported into the SUMO traffic simulator to be used as a method for calculating the fitness of solutions. Our results are competitive compared to the common solutions from experts in terms of travel and stop time, and also with respect to other similar proposals but with the added value of solving a real, big instance.
http://dx.doi.org/10.1145/2463372.2463540
INDIAN GCP GUIDELINE. for Regulatory affair 1st sem CRR
Red Swarm: Smart Mobility in Cities with EAs (GECCO'13)
1. Red Swarm:
Smart Mobility in Cities with EAs
Daniel H. Stolfi and Enrique Alba
Amsterdam, The Netherlands
July 06-10, 2013
2. RReedd Red SSwwaarrmm:: Swarm: SSmmaarrtt MMoobbiilliittyy iinn CCiittiieess wwiitthh EEAAss
Daniel H. Stolfi
dhstolfi@lcc.uma.es
http://www.danielstolfi.com
Enrique Alba
eat@lcc.uma.es
http://www.lcc.uma.es/~eat
University of Malaga
University of Malaga
July 2013
July 2013
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs
3. TTaabbllee ooff CCoonntteennttss
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 3
3
Introduction
Proposal
Experiments
Conclusions
1
2
4
4. Introduction
Proposal
Experiments
Conclusions
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 4
IInnttrroodduuccttiioonn
People are living or thinking about moving to big cities
A higher number of vehicles are moving by the streets
The number of traffic Jams is rising
Tons of greenhouse effect gases are emitted to the atmosphere
The citizens' quality of life is decreasing
The optimization of
road traffic is
necessary in modern
big cities
5. Introduction
Proposal
Experiments
Conclusions
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 5
RReedd SSwwaarrmm
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Our Red Swarm consists of:
Several spots distributed throughout the city
Installed at traffic lights
Linked to vehicles by using Wi-Fi
Our Optimization Algorithm
Our Rerouting Algorithm
On Board Units (OBU)
Installed inside vehicles
Smartphones or tablets
could be used instead
6. produces
CENTRALIZED OFFLINE ONLINE DISTRIBUTED
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 6
AArrcchhiitteeccttuurree
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
The Evolutionary Algorithm produces a configuration for the Red Swarm spots
The configured Red Swarm spots are deployed in several junctions of the city
7. RReerroouuttiinngg ((tt == 1100 ss..))
Introduction
Proposal
Experiments
Conclusions
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Experts' solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 7
8. RReerroouuttiinngg ((tt == 1111 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 8
9. RReerroouuttiinngg ((tt == 1122 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 9
10. RReerroouuttiinngg ((tt == 1133 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 10
11. RReerroouuttiinngg ((tt == 1144 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 11
12. RReerroouuttiinngg ((tt == 1155 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 12
13. RReerroouuttiinngg ((tt == 1166 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 13
14. RReerroouuttiinngg ((tt == 1177 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 14
15. RReerroouuttiinngg ((tt == 1188 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 15
16. RReerroouuttiinngg ((tt == 1199 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 16
17. RReerroouuttiinngg ((tt == 2200 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 17
18. RReerroouuttiinngg ((tt == 2211 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 18
19. RReerroouuttiinngg ((tt == 2222 ss..))
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Experts’ solution Red Swam
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 19
20. Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 20
21. Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Constructing tthhee SScceennaarriioo:: MMeetthhooddoollooggyy
We have worked with real maps imported from OpenStreetMap
We have added 10 Red Swarm spots in junctions controlled by a traffic light
We have imported the map into SUMO
We have defined traffic flows for vehicles (experts' solution)
This process can be adapted to any modern city in the world
NETCONVERT and DUAROUTER
are part of the SUMO package
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 21
22. SScceennaarriioo:: MMaallaaggaa
MALAGA (SPAIN)
Real Scenario
261 traffic lights
10 Red Swarm spots
800 vehicles
4 vehicle types
3 different traffic patterns
( Scen1, Scen2 & Scen3 )
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Sedan Van Wagon Transport
Our goal is to reduce the travel time of the vehicles
in high density conditions
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 22
23. Experiments
Conclusions
EEvvoolluuttiioonnaarryy AAllggoorriitthhmm
Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
(10+2)-EA
The fitness value is calculated by using the
SUMO traffic simulator
The rerouting produced by the Red Swarm spots
is implemented by using TraCI and the
Rerouting Algorithm which runs in each spot
The result of the EA is a configuration for all the
Red Swarm spots placed in the city
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 23
24. Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Evolutionary Algorithm: RReepprreesseennttaattiioonn ((SSeennssoorrss))
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 24
SENSORS
They represent the inputs to Red Swarm spots
in the simulated scenario
Vehicles trigger the rerouting algorithm when
they are detected by a sensor
In a real city they would be radio links
25. Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Evolutionary Algorithm: RReepprreesseennttaattiioonn ((EExxaammppllee))
For example, if a vehicle which is going to Destination 2 is detected by Sensor 1,
one of the routes from R121 to R12K will be chosen by the Rerouting Algorithm,
depending on the values of the probabilities P121 to P12K
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 25
26. Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Evolutionary Algorithm: RReepprreesseennttaattiioonn
REPRESENTATION
RSDK: Available routes from Sensors to Destinations and to other spots
Each new route will be selected by the Rerouting Algorithm depending on the
probabilities
Conceptual representation
Solution vector made up of 1119 floats (probabilities)
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 26
27. Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Evolutionary Algorithm: FFiittnneessss FFuunnccttiioonn
FITNESS FUNCTION
N: Total number of vehicles (800 in this work)
ntrips: Number of vehicles which have ended their itineraries
ttrip: Trip time of each vehicle
tdelay: Waiting time of each vehicle before entering the analyzed zone
The three terms are weighted by α1, α2, and α3, respectively
We want to minimize the fitness value (the lower the better)
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 27
28. Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Evolutionary Algorithm: RReeccoommbbiinnaattiioonn OOppeerraattoorr
RECOMBINATION OPERATOR
We use the standard two-point recombination
The offspring are obtained by exchanging a range of sensor
configurations from the selected parents
TPX
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 28
29. Red Swarm
Architecture
Rerouting
Scenario
Evolutionary Algorithm
Introduction
Proposal
Experiments
Conclusions
Evolutionary Algorithm: MMuuttaattiioonn OOppeerraattoorrss
MUTATION OPERATOR
Changes the probabilities of the routes in the sensor configurations
1) All Destinations – One Sensor
It changes all probabilities in a Sensor block (i.e. all the probabilities in Sensor 4)
2) One Destination – One Sensor
It changes probabilities in a Destination block in a Sensor block
(i.e. only the probabilities of Destination 8 in Sensor 4)
Exploration
Exploitation
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 29
30. Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 30
RReessuullttss
Introduction Results
Proposal
Experiments
Conclusions
The 800 vehicles leave the city in a lower time when we use Red Swarm
Results show a reduction of
the average waiting time and
the average travel time
Vehicles have to travel
a longer distance because
they are using alternative routes
31. Introduction Results
Proposal
Experiments
Conclusions
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 31
32. Introduction Results
Proposal
Experiments
Conclusions
Results: Simulation Time vvss.. NNuummbbeerr ooff VVeehhiicclleess
Figures show that the higher the number of vehicles is, the more
effective Red Swarm becomes
Under the threshold, Red Swarm still being useful
Advantages are evident when the traffic density raises
Travel Time (Scen1) Travel Time (Scen2) Travel Time (Scen3)
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 32
33. Introduction Results
Proposal
Experiments
Conclusions
GGeenneerraalliizzaattiioonn ooff RReessuullttss
This figure shows the fitness values of the execution of Red Swarm in
30 extra scenarios compared with the experts' solution
Red Swarm has not only worked in all these scenarios but has also
achieved lower travel times in 20 of them (66.7%)
Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 33
34. Daniel H. Stolfi and Enrique Alba Red Swarm: Smart Mobility in Cities with EAs 34
CCoonncclluussiioonnss
Introduction
Proposal
Experiments
Conclusions
This is an innovative approach to the prevention of traffic jams
The results confirm road traffic can be improved by using Red Swarm
We have reduced the travel times and waiting times of the vehicles
Currently, we are working on:
The expansion of the analyzed region
Collecting information from vehicles (on-line and historical data of the city)
The reduction of greenhouse gas emissions