Event Coverage in Theme Parks Using Wireless Sensor Networks with Mobile Sinks

Transcription

1 Event Coverage in Theme Parks Using Wireless Sensor Networks with Mobile Sinks Gürkan Solmaz and Damla Turgut Department of Electrical Engineering and Computer Science University of Central Florida - Orlando, FL June 10, 2013 G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

2 1 Introduction G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

3 1 Introduction 2 Event Coverage Network model Dynamic directed graph model Event coverage strategies G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

4 1 Introduction 2 Event Coverage Network model Dynamic directed graph model Event coverage strategies 3 Simulation Study Simulation environment Simulation results G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

5 1 Introduction 2 Event Coverage Network model Dynamic directed graph model Event coverage strategies 3 Simulation Study Simulation environment Simulation results 4 Conclusion G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

6 Introduction Theme parks are very large, bordered and crowded areas (e.g. Disney World in Orlando) Consist of separate various attraction areas distributed in different locations and walking paths Visitors are usually walkers without vehicles Modeling of human mobility allows us to simulate different types of networks Current use of networks: Data collection from visitors for commercial purposes, entertainment of customers, crowd management, and so on G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

7 Motivation & contribution Event handling and coverage is one of the major challenges in theme parks due to inevitable security and emergency problems The use of WSNs may help effective handling of security events such as pick-pocketing and managing emergency incidents We propose new strategies for mobile sink positioning and event handling decision problems to cover the events For sink positioning: Weighted, adaptive For event handling decision: Shortest path, closest sink G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

8 Network model Network consists of static sensor nodes randomly distributed and mobile sinks Sensor nodes have the ability to sense an event from their environment and send it to the closest sink There are multiple but limited number of mobile sinks with the ability to move quickly and handle the events Sinks collect data from the sensor nodes and share the information about their locations, computation results or event information among each other G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

9 Network model A mobile sink may be a security officer driving personal electric transportation vehicle (e.g. Segway Patroller) Devices with computation and wireless communication capabilities such as a tablet computer attached to the vehicle The speed of a mobile sink changes dynamically since the environment includes physical constraints (e.g. people, buildings) We estimate a speed between two locations for a mobile sink using a simple micro-mobility approach G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

10 Network model The events represent the security and emergency events which security officers should act The events usually occur inside or close to an attraction More people in an attraction causes an increase in the possibility of occurrence of an event Each event has a timeout which changes the mode of the event from active to passive If a mobile sink can move to the location in active time, the sink handles the event If the mobile sink is unable to arrive at the event location in a timely manner, the event is considered as missed G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

11 Human mobility model We use the scenario-specific human mobility model for modeling the attractions and the mobility of the visitors This mobility model synthetically creates the attractions using different types of queueing models The queues are created using the fractal points generated as visiting points DBScan algorithm is used for clustering the visiting points to find the most-populated areas people gather around G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

12 Dynamic directed graph model Dynamic directed graphs are used to model the environment The nodes represent attractions, mobile sinks, and events The edges are the roads for movement of the mobile sink Some pair of nodes do not have an edge connecting them, meaning that there is no direct connection For instance, an edge from an event to a mobile sink or an attraction is not necessary G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

13 Dynamic directed graph model M1 A1 A3 M2 A5 A2 A4 E1 G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

14 Dynamic directed graph model In the algorithm for edge weights, the number and directions of visitors, distance, and width of a road effect the movement time of a sink If the density of people is high then travel time is also high The edge weights change dynamically and updated according to the current movement patterns and positions of visitors in the park 12.5 mph is used as the maximum mobile sink speed as it is listed in the Segway Patroller specification for the models x2 and i2 G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

15 Event coverage strategies The problem of the event coverage is the capability of the mobile sink reaching an event sensed by sensor nodes in minimal time To achieve this goal, we need to position the mobile sinks select the best path and the best mobile sink The strategies include Weighted sink positioning Adaptive sink positioning Shortest path strategy Closest sink strategy G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

16 Weighted sink positioning For the problem of positioning mobile sinks, we first propose a weighted sink positioning strategy We assume that the number of visitors is proportional to the number of events occurring in the attraction By monitoring the approximate number of visitors of all attractions, we assign the probability values to attractions Update the probability values iteratively at discrete update times The mobile sinks are distributed to the attractions according to the current capacities G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

17 Adaptive sink positioning There may be a subset of attractions having higher rates of events compared to the other attractions This behavior seems probable since the theme park environments are dynamic and likely to have unexpected changes As a result, we propose another strategy, the adaptive sink positioning This time the mobile sinks keep history of all events happened by sharing the event data among each other The data includes the index of the closest attraction node and priority of the event G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

18 Adaptive sink positioning After occurrence of a new event, the probability of the closest attraction increases by a value The capacities of other attractions decrease by the same value divided by the total number of the other attractions The sinks have common location update times Each sink decides and moves to the next attraction at the update times according to the current capacities and occupancy by other sinks G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

19 Shortest path strategy For effective event handling, we apply the Shortest Path strategy The values are updated according to the current edge weights for each mobile sinks Each mobile sink finds its shortest path (travel time) to the event The sinks share their estimated travel time values among each other The sink with the minimum estimated travel time is selected as the best sink G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

20 Closest sink strategy When monitoring movements of visitors is not possible, we use the Closest Sink strategy We use distances instead of travel time estimates The shortest path is found by the static edge weights (distances) The edge weights are the Euclidean distance between the attractions G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

21 Simulation environment We use theme park (TP) and SLAW mobility models to model the human movements There are 15 attractions and one or multiple (up to 10) mobile sink nodes Edges between the attraction nodes are created in two ways: graph density of 0.7 and vertex degree of 4 Temporary edges connect the mobile sinks and the events to the closest attractions G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

22 Simulation environment Location update time for adaptive mobile sinks is 30 minutes. Events are created temporarily and distributed in three ways: Random selection Selection according to the attraction weights Selection from a number of randomly selected specific attractions G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

23 Results: Weighted sink positioning Average event handling time (sec) Random TP Weighted TP Random SLAW Weighted SLAW 0 Event coverage strategies with mobility models For both mobility models, the weighted sink positioning is a clear winner compared to the random positioning since the sinks need less time to handle an event G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

24 Results: Adaptive sink positioning Average event handling time (sec) Adaptive TP Weighted TP Adaptive SLAW Weighted SLAW Simulation run In adaptive positioning, no matter where the sinks are located initially, they are able to move Events may occur close to the attractions with low capacities in one experiment G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

25 Results: Event handling decision Hit ratio (%) SP TP CS TP RS TP SP SLAW CS SLAW RS SLAW Simulation run We place sinks randomly in Random sink (RS) strategy Closest sink (CS) strategy produces significantly better results compared to RS strategy Shortest path (SP) strategy always has the significant advantage over the other two strategies G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

26 Results: Use of multiple mobile sinks Theme Park SLAW 70 Hit ratio (%) Number of mobile sinks Sinks are expensive elements of the network, we expect to reach a convergence point Using multiple mobile sinks increased the success rate from 30% to 90% G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24

27 Conclusion In this paper, we tackled the event coverage problem in theme parks We proposed new strategies for event handling and sink positioning The success of our approaches are evaluated through different scenarios using the scenario-specific TP and SLAW mobility models G. Solmaz, D. Turgut (UCF) IEEE ICC 2013 June 10, / 24