Sonia Pinto ALL RIGHTS RESERVED

Transcription

1 2011 Sonia Pinto ALL RIGHTS RESERVED

2 A RESERVATION BASED PARKING LOT SYSTEM TO MAXIMIZE OCCUPANCY AND REVENUE by SONIA PREETI PINTO A thesis submitted to the Graduate School-New Brunswick Rutgers, The State University of New Jersey in partial fulfillment of the requirements for the degree of Master of Science Graduate Program in Electrical & Computer Engineering written under the direction of Professor Ivan Marsic and approved by New Brunswick, New Jersey October, 2011

3 ABSTRACT OF THE THESIS A Reservation Based Parking Lot System to Maximize Occupancy and Revenue By SONIA PREETI PINTO Thesis Director: Professor Ivan Marsic Numerous parking issues are faced by people on a day to day basis. Congestion is caused by customers circling the lot in search of vacant parking spots while the lot may be full. At peak periods the parking lot may not be filled to maximum capacity because there are only primitive ways to indicate to customers the availability of spots in the lot. Thus parking garages tend to lose profit as usage of parking lots is not maximized. The Parking Lot System proposed aims to reduce these parking hassles faced by people and the occupancy issues faced by parking lot owners by providing customers with a facility of reserving parking spots before arrival at the lot. This reduces the parking search traffic as well as the parking search time. Additionally, this system proposes to automate occupancy tracking in the lot which ensures that customers are aware of the availability of parking spaces from remote locations. In case of unavailability of parking spaces, customers are notified at the entrance thereby reducing congestion levels in the lot. ii

4 The designed system is simulated in real world scenarios to test the occupancy levels achieved by the lot. The practices and strategies used in well developed reservation systems (hotels and aviation industries) like overbooking are modified and applied in this domain to improve occupancy levels. An increase of about 5.19% in the occupancy level is observed. A comparative study of the effect of various business strategies is carried out to determine the best mix of user controllable parameters. The parameters controllable by the parking lot owners are the length of the grace period provided to customers to arrive and claim their reservation, the number of overbooked reservations made by the system and the proportion of spots in the lot available for reservation. A further increase of up to 5.97% in the occupancy level is observed by using different combinations and values of these parameters. For the purpose of simulation, the parking garage is assumed to be located in a city downtown area having customers using it for both corporate as well as leisure purpose. iii

5 Dedication To my Parents iv

6 Acknowledgement I would like to thank my advisor, Dr Ivan Marsic, for giving me an opportunity to work on this very challenging and interesting real world problem. I would like to thank him for his constant guidance and valuable suggestions throughout the course of my thesis. I would also like to thank my parents for their encouragement and understanding without which all this would not have been possible. It is their constant support and love over the years that has got me to where I am today. My time here at Rutgers has been marked by the friends I made along the way. I would like thank these friends for their help over the last 2 years. Last but not the least, I would like thank my roommates for their constant support, help and company during the two years of my Master s. They were catalyst in making this entire experience an enjoyable, fruitful and memorable one for me. v

7 Contents ABSTRACT OF THE THESIS... ii Dedication... iv Acknowledgement... v List of Tables... viii List of Illustrations... ix Chapter 1: Introduction Motivation Scope Assumptions Business Strategies Problem Statement Thesis Organization Chapter 2: Related Work Parking Automation Systems Overbooking Simulation Models Research Questions Chapter 3: Technical Approach Overall Architecture and Design of Smart Parking Lot Algorithms used to handle reservations and parking Distance Calculation Algorithms for Reservation Reserving a spot when an exact match is not found Finding optimal slots: The Parking Lot Usage Bitmap Simulator Design Chapter 4: Results Fuzzy Matching Algorithm Results Simulation Results Research Questions Discussions Chapter 5: Conclusion and Future Work Conclusion vi

8 5.2 Future Work Chapter 6: References vii

9 List of Tables Table 1: Summary of Simulation Parameters Table 2: Occupancy Levels achieved in Ideal Scenario Table 3: Occupancy Levels achieved in Real World Scenario Table 4: Occupancy Levels achieved in Real World Scenario with Overbooking Table 5: Variation of Percentage of Occupancy of the lot with Grace Period Table 6: Variation of Percentage of Occupancy of the lot with Reservation Limit Table 7: Variation of Percentage of Occupancy of the lot with Overbooking Offset Table 8: Variation of Percentage of Occupancy of the lot with varying Grace Period and Reservation Limit over constant Overbooking Offset of 0.78 Table 9: Variation of percentage of occupancy of the lot with varying grace period and reservation limit over constant overbooking offset of 2.0 Table 10: Variation of Percentage of Occupancy of the lot with varying Grace Period and Overbooking Offset over constant Reservation Limit of 100% Table 11: Variation of Percentage of Occupancy of the lot with varying Grace Period and Overbooking Offset over constant Reservation Limit of 80% viii

10 List of Illustrations Figure 1: System Architecture Block Diagram Figure 2: State Diagram of a Parking Spot Figure 3: Modified State diagram of a Parking Spot Figure 4: Division of Parking Lot into Areas for Dijkstra s Algorithm Figure 5: Directed Graph of Parking Lot Figure 6: The Parking Lot Usage Bitmap Figure 7: Simulator Architecture Block Diagram Figure 8: Reservations mapped to time slots in the test scenario for Fuzzy Matching Figure 9: Plot of Percentage of Occupancy against Time Intervals for Ideal case Figure 10: Plot of Percentage of Occupancy against Time Intervals for Real World Scenario Figure 11: Plot of Percentage of Occupancy against Time Intervals for comparison between Ideal Scenario and Real World Scenario Figure 12: Plot of percentage of occupancy against Time Intervals for Real World Scenario with Overbooking Figure 13: Plot of percentage of occupancy against Time Intervals for comparison between Ideal Scenario and Real World with Overbooking Scenario Figure 14: Plot of Revenue against α for variation of Grace Period Figure 15: Plot of Revenue against α for variation of Reservation Limit Figure 16: Plot of Revenue against α for variation of Overbooking Offset. ix

11 1 Chapter 1: Introduction 1.1 Motivation With the explosion in the number of cars in many densely populated areas, a common problem that has arisen is that of parking. A number of people experience the annoyance of not finding adequate parking near the restaurant or movie theater of their choice, or a good parking spot in a commuter lot during rush hour. Wouldn t it be a lot nicer if parking also could be planned ahead for? Wouldn t it be so relieving to plan a dinner, have the required reservation made at a restaurant and also have a parking spot waiting for us upon our arrival at the destination? As parking becomes a very essential requirement in our day to day life, we look forward to planning and acquiring a secured parking spot before heading out towards our destination in order to reduce the hassle of driving around looking for a parking spot during park hours. There are numerous multi level parking decks and large parking lots being constructed in big cities to address this problem of parking. As the volume of motorists looking to park increases, we tend to see a lot of congestion in parking lots and decks as people circle around in the lot or compete with one another in their attempts to find the closest and most convenient parking spot. A system which would avoid this congestion and the hunt for parking spots in large parking lots and multi level decks is desired in big cities. Wouldn t it be a lot nicer to head right away to one of the available parking spot upon your arrival at the lot thus avoiding the congestion?

12 2 Most of the parking garages currently lack any automation or a computerized system to keep track of occupancy. Occupancy is tracked manually and is thus prone to human errors. The parking garage owners maintain the garage by entering data into excel type programs. In most parking lots people are allowed to park as long as they want to and they pay for their parking at an hourly rate. People either pay at exits or at pay machines located inside the parking lot. Hence at peak hours the exits of the parking lot and pay machines tend to have long lines of people waiting to pay for their parking. Moreover, at peak hours when the parking lot may be full there is no way to indicate to the customers that there are currently no vacant spots available. Customers end up circling the lot in search of a vacant spot causing congestion in the parking lot. Such a parking lot is likely to face many issues in terms of accuracy as well as speed and efficiency of the system. These inefficiencies of the system would not be favorable towards encouraging customers to return to the lot. Additionally, at peak hours the parking lot may not be filled to its maximum capacity since there are only primitive ways to indicate to the customer the availability of spots in the lot. Thus the parking lot owners end up losing profit [10] [6] due to the lost opportunity of not maximizing the usage of the available parking spots. Since there is no way to indicate to customers remotely whether there are available spots in the lot at a certain time, customers may be discouraged to drive to the parking lot. Such customers add to the loss of revenue generated from the lot.

13 3 1.2 Scope The aim of this system would be to develop a solution to the parking hassles mentioned in Section 1.1 and to improve the efficiency and usability of parking garages thereby improving the revenue generated for parking lot owners. Such a system would provide customers with a facility to plan for their parking before arriving at their destination. This facility is provided by introducing the concept of reservations in parking lots which gives customers the privilege of reserving parking spots before their arrival at the lot just like reserving a table at a restaurant or booking a hotel room. The reservation algorithms used aim to reduce congestion levels in large parking lots (200 to 700 spots) and garages and hence reserve spots which are most easily accessible to users. This feature ensures that there is less congestion caused by customers with reservations looking for their assigned spots thus minimizing the time spent by the customers in the parking lot. This system aims to provide predictable and guaranteed parking to customers with reservations and also serves the general public with no reservations. The reservation algorithms are explained in detail in the Technical Approach Section of this document. Additionally, it is essential that these large parking lots and multi level parking decks are well used by the general public as a lot of money is invested in making them. This is achieved by automating the parking garage from occupancy monitoring to payments in order to improve the speed and efficiency of the garage. The status of each parking space is monitored using sensors which enables customers to remotely check the availability of parking spots in the lot encouraging customers to use the vacant parking spots. If the parking lot has no vacant spots, customers are informed of the unavailability at the

14 4 entrance. This reduces the congestion level in the lot caused due to traffic searching parking. Customers with reservations can make payments through their credit card and do not require access to the payments machines in the garage. This improves the customer level service and efficiency of the system encouraging customers to return to the parking lot. The pay and park policy used by the system gives a good approximation of the current and future occupancy of the parking lot. This approximation helps the system plan reservations in order to avoid conflicts with existing parkings and reservations. Since the implementation of the proposed system in a parking lot requires certain hardware components such as sensors which currently not available in most parking lots, it was not possible to build and implement this Smart Parking Lot system in an existing parking lot. So, a Parking Lot Simulator is used to evaluate the proposed system and test the efficiency and occupancy levels achieved by it. The system is simulated taking into consideration the real-world events that would affect the system s ability to honor reservations and maintain high occupancy levels at the same time. These real world events include late arrivals and no shows among others. Various probability distribution models are used to simulate the arrival and departure of customer, along with the simulation of the duration of time that different customers park for. The details of the simulator and the probability distribution models are explained in the Technical Approach section of this document. An attempt is made to improve the achieved occupancy levels by introducing overbooking. Overbooking is widely used and accepted in other industries such as hotels

15 5 and aviation which have well defined and efficient reservation systems. Overbooking refers to accepting more reservations than the system can actually honor in order to protect the parking lot owners from losses caused due to vacancies in the parking lot due to late arrivals and no-shows. Overbooking techniques used in the hotel and aviation industry make use of their historical data to determine the acceptable level of overbooking that their reservation system should go up to. But since reservation systems in the Parking industry are a novel idea, it was not possible to get such historical data for our calculations. Hence we estimate a reasonable percentage of no shows, early departures and late arrivals in order to implement overbooking strategies. These assumptions are clearly listed in the assumptions section below. The overbooking algorithm implemented is explained in detail in the Technical Approach section of this document. The increase in the occupancy level achieved through overbooking is shown in the Results section of this document. The parking lot designed uses various business strategies in order to achieve high occupancy levels. These business strategies are based on parameters in the proposed system which can be controlled by the parking lot owners to make sure to achieve high occupancy in the face of changing external factors. The changing external factors and the parameters that can be controlled by the parking lot owners are as follows: The external factors that affect our system s performance are: Rate of arrivals of customers and rate at which bookings are made. Average parking time (reservations and walk-in customers).

16 6 Percentage of customers that are delayed in arriving and departing. Amount of time by which customers extend their parking. Fraction of customers that choose to extend their parking time. Delay in customer arrivals and departures (for customers with reservations). Percentage of no-shows. The parameters that the owner can control to react to these changes are: Length of Grace period for reservations Reservation Limit that determines the proportion of garage capacity allocated for reservations Overbooking offset that controls the amount of overbooking made Grace period is the time given to customers with reservation to claim their reserved spots and to vacate the spot without being penalized. Reservation Limit is the percentage of spots available for reservation in the parking lot. The remaining percentage of spots would be available only to walk-in customers. Overbooking offset is a factor used by the overbooking algorithm which defines the number of extra reservations that the system accepts. More details on these user controllable parameters are discussed in Section 4.2. Different parameters values are used and simulations are run to evaluate the effect of varying these parameter values individually and in combinations. The results are studied very closely in order to evaluate the best business strategies to be applied in order to achieve high occupancy levels. These results are tabulated in the Results section of the document.

17 7 Thus by using algorithms and business strategies the Smart Parking Lot system aims to reduce the parking hassles faced by people and at the same time tries to maximize the profit generated for the parking lot owners through business strategies. The assumptions and business strategies adopted by this system are explained below the following two sections. 1.3 Assumptions The Smart Parking Lot System designed is based on the following assumptions: 1. The parking lot is situated in a city downtown area and has customers using the lot for both corporate as well as leisure purposes. 2. The parking lot has two entrances/exits available to customers to enter and exit the lot from. 3. The parking lot consists of 4 levels 200 parking spots with 50 spots on each level. 4. Corporate customers park for a longer duration of time for about 8 to 9 hours. Customers who park overnight also park for similar long hours. 5. Customers who use the lot for leisure purposes park in the lot for shorter duration of time for about 3 to 5 hours. 6. For the purpose of overbooking we assume that 15% of the customers who have reservations do not show up (in close accordance to hotel and aviation industries). 7. Users are allowed to book (and also park) for multiples of time-slots of a fixed duration. The length of the time slot is chosen as 15 minutes. This is similar to many metered parking systems (e.g. where one gets 15 minutes of parking time

18 8 for every quarter inserted into the meter). Using such a system streamlines many processes related to parking, booking and making payments. 8. For simulation purposes, the database snapshot is taken once every 7 minutes to determine the state of the system i.e. number of reservations, number of free spots etc. 9. The system is designed in terms of hours and minutes keeping in mind the real world scenario but for the sake of simulations we scale down hours to minutes and minutes to seconds in order to complete the simulations within a reasonable amount of time. 10. Walk-in customer arrival rate (lambda) is assumed to be 40 per hour in the real world scenario. The remaining 160 spots are available for reservations. The initial working inventory for the overbooking algorithm is assumed to be The rate of arrival of reservations (lambda) is assumed to be 60 per hour. 12. The customer service level is assumed to be 1.28 [30]. Overbooking offset is Business Strategies The Smart Parking Lot system designed is based off the following business practices: 1. The Smart Parking Lot proposed uses the Pay and Park policy. This means that the customers are required to pay for their parking spots immediately after they have parked their vehicles for the amount of time they are expected to be parked. This is a pre-paid parking lot.

19 9 2. The parking lot is open to both customers with reservations as well as customers with no reservations. Customers with no reservations are termed as walk-in customers. 3. Customers with reservation are given priority over walk-in customers. At all times it is highly essential to honor reservations in order to maintain a high customer service level in order to ensure that customers return to the lot. 4. The cost of parking for customers with reservations is slightly higher than the cost of parking for walk-in customers. 5. Customers who use the lot for corporate purposes usually have a reservation pass valid for a certain period of time say for example 1 month. They are considered to be customers with reservation as a spot is reserved for them. 6. Customers who wish to make reservations with the system must be registered on the Web Application and provide details such as a valid address and credit card number. 7. When a customer makes a reservation he is given a unique confirmation code. He is not provided with the spot number reserved during reservation. 8. The parking lot is completely automated from entrance for monitoring payments. 9. There is an Entry Console at each of the two entrances to the lot. This console requires users to indicate whether or not they have reservations. If they have reservations their spot number is provided to them by the console after they key in their unique confirmation code. If not, they are indicated to go park in any vacant spot. If there are no such available vacant spots in the lot they are informed of the unavailability in the lot and are not granted access to the lot.

20 Customers with reservations would pay for their parking online through their credit card. The payment would be charged on their card when they leave the parking lot. 11. Walk-in customers are required to enter their desired parking duration and pay for their parking at the Parking Time Entry Console once they have parked their vehicles. There are two Parking Time Entry Consoles located on each level. 12. Customers with reservation would be charged for their entire reserved time even though they may arrive later (within a grace period) than the start of their reservation time. 13. If customers with reservation overstay they would be charged heavily for the extra time they stay beyond the grace period. 14. Customers with reservations have the option to extend their parking duration. This extension can only be done a certain amount of time before the reservation end time (e.g. one hour). 15. Walk-in customers are given the privilege to park any in any vacant spot in the lot. 16. When the parking lot is full (taking into consideration both occupied as well as reserved spots) at the current time, walk-in customers are indicated at the entrance of the unavailability of spots in the lot and are not allowed to enter the lot. 17. Customers with reservation are given a grace time beyond which they can no longer claim their reservation. This grace period can range from 10 to 30 minutes and is dependent on the rate of arrival of customers. Customers with reservation can arrive after the start of their reservation time but within the grace time and

21 11 park in their reserved spot. If the customer arrives after the grace period he would have lost his spot and will be treated as a walk-in customer. He will be notified of this at the Entry Console on entering his confirmation number. The same is applicable with the end time of reservation as well. Customers can depart after the end time of their reservation and within the grace period provided to them in order to avoid any penalty. 18. If there are no spots available for the requested time slot when a customer tries to make a reservation, he is given an option to make a reservation for other closely matched time slots, though not the exact time slot or to overbook. Overbooking is done in the system to compensate for customers who depart early or not show up at all. When a customer makes a reservation which would result in overbooking the lot, he is informed while making such a reservation that he may not be able to get his reserved spot because the parking lot may be full. 19. Customers are penalized heavily for overstaying beyond the stipulated acceptable time (grace time). Registered users can be charged a fine on their credit cards, while walk-in customers could be ticketed. 1.5 Problem Statement This section identifies the problems of parking we are trying to tackle in this system. There are two major issues we are trying to find a solution to. The first one is solving the parking hassles faced by people. The second one is the inefficiency in the current parking garages. The approach we take to solve these two issues and evaluate our solution is explained below.

22 12 The parking hassles faced by customers as discussed in Section 1.1 are resolved by providing a parking lot that can make reservations, as well as be available to walk-in customers. Reservations would cut down the number of people driving around to look for a vacant parking spot, which causes 30% of the traffic in big cities [19]. Also, assigning spots which are most readily accessible by customers on entering the lot would reduce the congestion caused by customers driving around the lot looking for a vacant parking spot. Parking garages should also serve walk-in customers because such customers represent immediate revenue. In such a system, some customers are likely to park in spots not reserved for them. The system should then be able to dynamically move around reservations in the lot in a transparent manner, so that all reservations are honored. Business strategies must also be developed to stop new walk-in customers from entering when it becomes impossible to re-schedule reservations to honor them. A grace period should be provided to customers so that their reservation is held for a reasonable amount of time if they are late in arriving. But holding a spot for too long causes it to lie vacant, when it could be used by someone else. The optimal grace period needs to be determined through simulation on the system. Also, it is likely that a number of customers who have made reservations may not show up at all. To protect the system from losses caused by these no-shows, the system needs to implement an overbooking algorithm that can improve occupancy in case of no-shows without having to make a large number of reservations and later be unable to honor them. Overbooking techniques used by the hotel and aviation industries are used to deal with this issue. We study the arrival and departure patterns in these industries and try to draw parallels to the Parking Lot system.

23 13 Having a customer to park in the garage one at a time may not generate interesting behavior to analyze the stability of the system and to determine whether the designed business strategies would work to generate good revenue and usage of the system in the real world. Also it would be hard to simulate many users accessing the parking lot simultaneously in close accordance with the real world scenario. Hence it is required to develop a simulation model on the server side that can handle simultaneous interactions using which we can simulate the Parking Lot System in close accordance with the real world scenarios. The simulation model would simulate all the real world scenarios in relation to a Parking Garage. These scenarios include arrivals, departures, no shows, reservations, customers arriving late, customers departing late, extending parking durations, amount of parking time and other defaults as well. We can model the different customer actions using different probabilistic models. Simulating the system as close as possible to the real world scenarios is very important to study the efficiency of the business strategies designed and the efficiency of the algorithms used to reserve parking spots. 1.6 Thesis Organization The rest of the document is organized as follows. Section 2 gives an overview of existing solutions to tackle this problem of parking and ends with a few research questions which we try to find a solution to in this thesis. Section 3 explains the system architecture, algorithms and approach chosen to tackle the problems listed in Section 1. Section 4 explains the results we achieved through simulations we ran on the system and how these

24 14 results reflect what we have achieved through the system. Section 5 concludes and explains the possible future work in this direction.

25 15 Chapter 2: Related Work The problem of parking and congestion in parking lots has been around since several years. Parking Guidance Information (PGI) systems were provided in the early years to minimize parking search traffic in large parking facilities and central cities by dynamically monitoring available parking, and directing motorists with changeable message signs (CMS) [24]. Parking Guidance Systems based on Wireless Sensor Network (WSN) were also designed for automated and accurate monitoring of the parking spots and guidance to a vacant parking spot [32]. However these systems do not provide the facility to secure a parking spot in one of these lots before arrival. They only provide information regarding availability in each lot. 2.1 Parking Automation Systems Several initiatives have been taken to automate parking lots and reduce levels of congestions due to parking search traffic and parking search time. Some of them have been discussed below. A Smart Parking Lot System field test was conducted by researchers from December 2004 to April 2006 [25] at the Bay Area Rapid Transit (BART) in San Francisco. Both internet and phone reservations were provided. The online reservations were available over ParkingCarma [22]. The ParkingCarma website does not give any information regarding the how reservations are made and how slots are associated to users. Users indicated that they preferred the online reservations system over the telephonic one and

26 16 would like to have more than the minimum (3) number of advanced reservations. This paper does not give any information regarding occupancy of the lot, how spots are assigned and what happens when a user parks is a spot reserved by another user. Another initiative towards these parking hassles was taken by five students who proposed and implemented a system called EzPark [5] in which customers at the entrance of the lot would get a hint where to park along with two RFID tokens. One of these two tokens would be left in the car and the other one would be taken by the user. The RFID token in the car is used to identify which spot the user is parked in and whether the spot is occupied or not. The entire parking lot is filled with RFID readers. This system solves the problem of congestion in the parking lot by providing hints to the customer on which spot to park in. But the system may give a hint to customer A about a spot which customer B may be headed towards. Then either of these customers will have to find a vacant spot in the lot in the primitive way or return to the entrance to get another hint. A wireless sensor network based system proposed [27] provides advance features such as remote parking monitoring, automated guidance and parking reservation mechanism to deal with parking hassles. The remote parking monitoring is achieved through the Wireless Sensor Network which is deployed in the parking lot. A Web Application is used by customers to reserve parking spots remotely. A prototype model of this system was implemented to conduct some preliminary tests which showed that this system can effectively solve the existing parking hassles. The architecture of this system is the source of inspiration for our system.. This proposal does not talk about how spots are assigned to customers with reservations. There is no mention about parking defaults and no shows too. Vehicular ad hoc networks are used to search for free parking spots. This application guides drivers to

27 17 a free parking lot at their destination and this requires the estimation of occupancy state of parking lots around the destination area at arrival time of the customers [4]. A prediction model is introduced to predict the availability based on information exchanged among vehicles. Another initiative describes a multi-hop wireless parking meter (PMNET) network which when coupled with a GPS receiver allows users to quickly locate and navigate to an existing available parking space [2]. Users with any mobile device or a car navigation system can communicate with the parking meters, book and spot and pay for parking even before arriving at the lot. But the same vacancies are displayed for all users, if the user would just want to check for vacant spots. This might cause multiple users to drive to the same spot and find it occupied when they get there. Another reservation protocol was designed using vehicular ad hoc networks where drivers can receive information about parking spaces around them while driving and the protocol used efficiently allocated parking spots to interested vehicles thereby avoiding competition between them to get to the spot [9]. The reservation protocol used makes sure that only the availability of a single spot for reservation is indicated to the user and a list of available spots is not given to him. But several vehicles leaving their parking spots can communicate to a user of the availability of a spot there by creating confusion about which spot the user should respond positively to. Multiple users may respond positively to the message sent by a single vehicle leaving the parking spot. Moreover users can confirm and not turn up at the spot. There was another initiative of a parking search system using Wireless Sensor Networks called ParkNet [20] in which two ultrasonic sensors are fitted on the side of the car which collects information about various spots as the car is driven through different areas. This data is sent over a wireless network to an

28 18 internet server. The system then merges the two sets of data and develops color coded maps that show where on-street parking is likely available and where it s not. This system was tested in the Highland Park, NJ area and it was 95% accurate. These systems give a unique solution to the problem of finding street parking when required. They deal with roadside parking challenges. One of the largest Parking Guidance System, Sipark [26], is in the making at the Munich Airport by Siemens where each individual spot is monitored using ultrasonic sensors. Cars are guided to vacant spots by LED displays all over the parking lot. The solution also includes zone and aisle counting. This means that any vehicles still en route to a parking space are also acquired by the system which avoids guiding too many drivers into a sector that only contained a few unoccupied spaces when the car entered the car park. This perhaps solves the problem of many other systems mentioned above which had multiple users heading towards the same parking spot. Using ultrasonic sensors to monitor the occupancy of each spot is an interesting initiative and our system is inspired by this way of tracking status of parking spots. All the initiatives discussed above concentrated more towards guiding a user to a parking spot to reduce parking search traffic. Parking Guidance system perhaps does help to reduce congestion in large parking lots by a considerable extent. But several issues such as cars not being assigned to a particular may create chaos in the lot leading to not solving the problem of congestion. None of these systems aimed at increasing the occupancy of the parking lot by encouraging reservations. A few systems spoke of automated entrance and payment.

29 Overbooking A common problem that all reservation systems need to deal with is that of loss of revenue due to customer defaults. Customers with reservations depart early, arrive late, may overstay and some may not show up at all. Overbooking is an essential part in all systems that handle reservations. Cancellation of reservations and no shows occur very frequently in reservations systems. Loss of revenue due to cancellations and no shows needs to be minimized to keep the business profitable. Hence overbooking is used to compensate for no shows and cancellations. Well established reservation systems like in the industries of aviation and hotels have developed overbooking strategies to overcome this loss of revenue due to customer defaults. This problem of overbooking has been successfully tackled in the hotel [15] [30] and airline industry [28] [14] change to new added ref. Overbooking is done by forecasting the number of no shows and cancellations. Hotels and airline industries use historical data to determine the average number of no shows, cancellations, overstays and under stays. Apart from well known reservation systems in hotels and aviation which adopt overbooking techniques, health clinics also make use of appointment overbooking in order to avoid reduction in productivity due to patients not showing up [17]. 2.3 Simulation Models The system designed requires us to simulate the arrivals and departures of customers into the parking lot. Such simulations have been carried out in the past in order to evaluate and test the efficiency of various systems. One such system is Downtown Space Reservation System (DSRS) which was one of the transportation strategy developed to

30 20 mitigate congestion is the center of a city [33]. In this system, customers who want to travel to a certain area in the downtown area will have to book their trip with the DSRS system beforehand. This research proposal builds a simulation model where the process of a vehicle entering the designated downtown network is assumed to be a Poisson s process with a known average rate lambda. Another research initiative to build a intelligent GPS based vehicle navigation system to find street parking lots also uses Poisson s process to model the occupancy/availability of the lot while testing the system [23]. This is a prediction based system which accepts the user s destination as input and calculates the approximate time taken by the user to reach the destination from his current location. It then calculates based on past statistics of occupancies and the current state of the parking lot, the probability of their availability at the time when the user reaches the destination and returns information about the lot which has maximum probability of availability to the user. Miami University used simulations to evaluate parking space availability with the current layout of parking lots and to determine future design options [11]. For this simulation, cars entering the parking lot followed a non homogenous Poisson s process with a mean arrival rate depending on the time of the day. A patient room simulator was built in one of the hospital research initiatives where patients are presented for admission to the hospital in a Poisson distribution [29]. Revenue Management Systems in hotels have also used Poisson s Process to model incoming future reservations for the purpose of simulating their revenue system [1].

31 Research Questions Several factors will determine how the reservations and walk-in customers are handled in the lot, and this in turn will affect the overall occupancy and the revenue generated by the parking lot. Some of the questions which arise from the factors to be considered are as follows: 1. How to maximize the number of reservations in the parking lot? 2. How to preserve and honor reservations made by users? 3. How to maximize the availability of the parking lot for customers that do not have reservations? 4. How much grace period to allow for reserved customers that show up late? 5. How to allow extensions to parking for customers? 6. How to manage overbooking and what to do if a requested reservation cannot be met? 7. What distribution models to use to simulate the real world scenarios of arrivals, amount of parking time, late arrivals, defaults, no shows and parking time extensions and departures. The answer to the above questions gives us a set of business strategies. These strategies are obtained by analyzing the results obtained by simulating the system, and then addressing the above questions using the results of this simulation. In the later sections, we describe the business strategies that help us make the most optimal utilization of the proposed parking system. The answers to these Research Questions are listed out in Section 4.3.

32 22 Chapter 3: Technical Approach This section discusses the technical approach chosen to tackle the hassles of parking, making and honoring reservations and improving overall efficiency of the lot. The algorithms used to reserve the most optimal spot for the user and to deal with defaults and no shows are also discussed in this section. 3.1 Overall Architecture and Design of Smart Parking Lot The system proposed keeps the reservation structure of the parking lot as transparent as possible to the users. Customers with reservations are only given a confirmation number when they make a reservation with the system. Walk-in customers are free to park in any vacant spot for most of the time, regardless of whether that spot holds a reservation or not. This system is responsible for re-allocating the reservations that may conflict with the newly parked customer. The benefit of this approach is that all spots are available to walk-in customers to park, and no spots go unused in the time that the customer who has made the booking arrives. Customers with reservations are given their assigned spot number upon arrival at the lot on keying in their confirmation number provided during reservation. This way the customer will get a spot for the duration he requested irrespective of any re-shuffling the reservations the parking lot may have gone through. The overall architecture and design of the system proposed (hardware and software) from automatic occupancy tracking to the client web application is as explained below.

33 23 The Smart Parking System proposed is a Wireless Sensor Network based system and is divided into four subsystems as discussed below. This thesis specifically concentrates on the design of the Server Subsystem. The design of this system is inspired from several other Wireless Sensor Network based parking systems [3] [16]. The high level design diagram of the components of thiss application is as shown below in Figure 1. Fig. Fig.1 System Architecture Block Diagram Wireless Sensor Networks Subsystem Automation of occupancy tracking could be achieved through Wireless sensor networks (WSN) which are used keep track of the status of each parking spot in the lot. Wireless Sensor networks are chosen because they can be deployed in existing car-parks car without

34 24 having to install new cabling for network and electricity to reach each sensing device. Additionally, they also have the flexibility to couple with sophisticated but cheap sensors that can accurately detect vehicles [27] [18]. A sensor node is deployed in each parking spot in the lot to monitor its status. All the sensor nodes communicate through a wireless channel. Each sensor node maintains the status of the parking lot to be available, reserved or occupied. Ultrasonic sensors are used as a part of our Wireless Sensor Networks [27] [18] because detection by them is easy and reliable. Ultrasonic sensors are non intrusive sensors and hence do not affect the surface in the process of mounting them. They need to be at direct line of sight to the parking spot and hence are mounted on the ceiling directly above each parking spot. These sensors are the ideal sensors to be used in multi-deck parking lots [31] and are also economical to use as a part of the Wireless Sensor Networks in the lot Gateway Subsystem The status of each of these sensor nodes would be collected by the gateway which delivers this information to the database on the Server Subsystem [27] [18]. The gateway subsystem acts as a central hub which collects information about all the sensors in the parking lot. The gateway subsystem is connected to the database (which is a part of the server subsystem) via the internet. Information from the database is relayed to the sensor nodes (and vice versa) to keep the system updated in real time.

35 Server Subsystem The server subsystem as shown in Figure 1 consists of the database, Parking Lot Application which consists of algorithms to handle and make reservation, park cars, vacate, fuzzy matching of timeslots, overbooking and the business strategies implemented in the Parking Lot system and the Simulator which simulates the real world interaction with the system. In this thesis we concentrate fully on this subsystem and design the reservation algorithms, database and evaluate various combinations of business strategies to achieve maximum occupancy of the lot. The database is used to store information regarding the status of each spot in the lot. Each spot can have one of these 3 states: Free, Reserved or Occupied. The status of each spot is determined by the sensor nodes at each spot and by the reservation algorithms. The status is then updated in the database in order to keep the entire system updated in real time. Each spot can have several reservations for different time intervals. When a spot is booked for any time interval, however short or long it may be, the reservation algorithms change the status of this spot in the database to reserved. When a car parks in a spot, its status changes to occupied and this is detected by the sensor node at the spot. When a car leaves the spot, its status changes to free or reserved depending on whether the spot has any reservations or not. In the real world system, the sensors deployed in each parking spot are responsible for this change of status of each spot when a car arrives, vacates and when reservations are made.

36 26 The change of status of a spot when a customer books a spot, arrives and departs can be indicated clearly through a state diagram as shown below in Figure Reserve a spot Reserved Free 2. Arrive and Park 3. Depart Occupied Fig.2 State Diagram of a Parking Spot When a currently occupied spot is reserved by another customer for a different time period, the status of that spot in the database should not change from occupied to reserved. Since the sensor at that spot can sense the presence of a vehicle, the status of that spot continues to remain occupied. When the vehicle currently parked departs from the spot (sensor node senses the vacancy of the spot), the reservations database is checked to see if there are any future reservations for that spot. If there are future reservations, the status of the spot goes to reserved else it goes back to free. This is done to make sure no overlapping reservations are made in the system.

37 27 These conditions in the change of status of a spot can be clearly depicted through a state diagram as shown below in Figure 3. Reserved 1. Car 1 Reserves Spot 1 Free 2. Car 1 Arrives and Parks 4. Car 1 departs 5. Car 2 Arrives and Parks 6. Car 2 departs Occupied 3. Car 2 reserves Spot 1 Fig. 3 Modified State diagram of a Parking Spot The database also stores all the reservations made by customers as well as all entries for the current walk-in customers until their expiry time (i.e. until the customer departs from the parking lot). This is necessary in order to keep track of the occupancy of each spot to make further reservations for customers apart from keeping track of all the reservations made in order to assign the spot reserved for the customer when he arrives. Apart from this, the database also stores the details of the users who have access to the online Parking Reservation Application. Details such as Credit Card Number and License Plate numbers are stored in order to automate the payment process for customers with reservations as well as to check whether the customer is parked in the correct spot.

38 28 As shown in Figure 1, the server subsystem also contains the Parking Lot Application which consists of algorithms to handle parking and reservation which are explained in detail in Section 3.2. The Undirected Shortest Distance algorithm and the Dijkstra s algorithm are used to reserve parking spots for customers. These algorithm take the entrance of the parking lot as an input to get the nearest available parking spot for the desired time period requested in order to reduce congestion in the parking lot. The user is provided with an option to indicate which entrance he would be entering the lot from during reservation. These reservation algorithms are explained in details in Section 3.3. The customer is not given the spot number assigned to him until he arrives at the lot and keys in the confirmation ID provided to him. The fuzzy matching of time slots and overbooking is adopted when exact match to user requirements is not found. These algorithms are explained in detail in Section 3.4 As described in the problem statement section, since the Parking Lot System is a reservation based system it is important to deal with wrong parking as well. By wrong parking we mean the following cases: walk-in customer parking in a spot which is reserved by another customer and customer with reservation parking in a wrong spot. When a walk-in customer parks in a spot which is reserved, the reservations for this spot should be transferred to another spot. The system senses this when the walk-in customer keys in the spot number in the Parking Time Entry Console and pays for the parking when the sensor node has sensed and updated the status of the spot to be occupied. If a customer who had reserved this spot arrives and parks at the spot, he would not go up to the Parking Entry Time Console to enter his hours and pay as he would have already

39 29 done so online. When the customer keys in the spot number at the console, the status of this spot is checked in the database. If there are reservations for this spot which conflicts with the existing parking then all the conflicting reservations are shifted to other favorable spots. The reservations are shifted to other favorable spots by using reservation algorithms. Once the conflicting reservations are rebooked, the confirmation numbers are then copied over from the old reservations to the new one before deleting the old reservation. This way the customer who has reserved a spot can still access his reservation by keying in the confirmation number at the entrance of the parking lot. This transparency in rebooking means that the customer with reservation can get a parking spot, and would have no idea that someone else had actually parked in the spot assigned to him initially. Walk-in customers are free to park in any spot. Customers with reservations can also park in wrong spots. Like for example, a customer is assigned spot number 10 and while the customer is on his way to 10 he notices that spot number 6 is free and he parks there. The sensor node would update the status of the spot number 6 to occupied. Since the customer already had a reservation, he would not walk up to the Parking Time Entry Console. In such a case, the reservations of the two spots will be swapped. Apart from this the server also consists of the simulator which simulates real world interaction with the lot. Details of this can be found in Section 3.6