GT CS 4440 - Bus Prediction Algorithm Evaluation

Unformatted text preview:

CS4440 CS4440 Emerging Database TechnologiesEmerging Database TechnologiesProject ProposalProject ProposalOctober 2, 2007October 2, 2007Bus Prediction AlgorithmBus Prediction Algorithm EvaluationEvaluationJohn AbrahamJohn AbrahamGalen HusseyGalen HusseyMatt Weber Matt Weber1. IntroductionProviding accurate and timely information about bus arrival is important for Intelligent Transportation Systems (ITS) which use Automatic Vehicle Location (AVL). It drastically enhances the user experience while at the same time providing the operators with information about lagging or leading buses which can then take corrective action. Our objective in this project will be to utilize historical and current GPS records of the Georgia Tech Bus system to implement and compare algorithms for better bus arrival-time prediction. 2. Related Work This project will attempt to build upon two existing projects, BuzzRoute and the SMS bus tracker. Here we briefly summarize both projects. We then discuss the related the current arrival time prediction service employed by GA Tech NextBus.BuzzRoute is a project to provide a visual interface to the bus locations on a mobile device. BuzzRoute itself was built on a previous prototype that successfully implemented the mobile client and architecture, improving the architecture in some places, and adding a route detection component. Much of the code used in BuzzRoute is relevant to our bus prediction, so portions will be re-used in our application.[6]SMS bus tracker was a project to provide an SMS-based query response interface to the NextBus (described later) predictions. Unfortunately, this project was not completed, but it has plenty of code relevant to our application. We hope to use this project as our interface to the bus prediction algorithm. To our knowledge, the only service that attempts to predict the arrival time of a Georgia Tech bus is NextBus, which uses the existing schedule of the bus system, as determined by the transportation directory, and simply predicts the arrival time as the scheduled time of arrival. This system works well in cases where a given bus is either running early or on schedule, because the drivers are required to delay their progress to allow the schedules to catch up. This system’s obvious flaw is in cases where the buses are running behind schedule, since there is little a bus driver can do to compensate for most causes of delay, such as traffic or large numbers of people entering or leaving the bus. The bus schedules are broken into blocks for each route. Each bus on a single route is given its own block, which has static times for its arrival at each stop. The tables below show samples of the block schedule for the Green route, effective August 2007. Block Lv. Lv. Lv Ar. Lv. Lv. Lv. Ar.1 TEP CRC Hemphill & 10th St. GTRI GTRI Hemphill & 10th St. S. Cntr TEPGarage 7:00 AM 7:15 AM 7:19 AM 7:21 AM 7:25 AM 7:30 AM 7:35 AM 7:38 AM 7:42 AM1 7:45 AM 7:49 AM 7:51 AM 7:55 AM 8:00 AM 8:05 AM 8:08 AM 8:12 AM1 8:15 AM 8:19 AM 8:21 AM 8:25 AM 8:30 AM 8:35 AM 8:38 AM 8:42 AM1 8:45 AM 8:49 AM 8:51 AM 8:55 AM 9:00 AM 9:05 AM 9:08 AM 9:12 AM1 9:15 AM 9:19 AM 9:21 AM 9:25 AM 9:30 AM 9:35 AM 9:38 AM 9:42 AM1 9:45 AM 9:49 AM 9:51 AM 9:55 AM 10:00 AM 10:05 AM 10:08 AM 10:12 AM1 10:15 AM 10:19 AM 10:21 AM 10:25 AM 10:30 AM 10:35 AM 10:38 AM 10:42 AM1 10:45 AM 10:49 AM 10:51 AM 10:55 AM 11:00 AM 11:05 AM 11:08 AM 11:12 AMBlock Lv. Lv. Lv Ar. Lv. Lv. Lv. Ar.2 TEP CRC Hemphill & 10th St. GTRI GTRI Hemphill & 10th St. S. Cntr TEPGarage 8:00 AM 8:15 AM 8:20 AM 8:23 AM 8:27 AM8:30 AM 8:34 AM 8:36 AM 8:40 AM 8:45 AM 8:50 AM 8:53 AM 8:57 AM2 9:00 AM 9:04 AM 9:06 AM 9:10 AM 9:15 AM 9:20 AM 9:23 AM 9:27 AM2 9:30 AM 9:34 AM 9:36 AM 9:40 AM 9:45 AM 9:50 AM 9:53 AM 9:57 AM2 10:00 AM 10:04 AM 10:06 AM 10:10 AM 10:15 AM 10:20 AM 10:23 AM 10:27 AM2 10:30 AM 10:34 AM 10:36 AM 10:40 AM 10:45 AM 10:50 AM 10:53 AM 10:57 AM2 11:00 AM 11:04 AM 11:06 AM 11:10 AM 11:15 AM 11:20 AM 11:23 AM 11:27 AM2 11:30 AM 11:34 AM 11:36 AM 11:40 AM 11:45 AM 11:50 AM 11:53 AM 11:57 AMAnother flaw NextBus suffers from is in the bus system itself. In order for the route to be reported from the bus correctly, the bus driver must enter the route code into a keypad interface. Due to intermittent hardware problems, this is sometimes not accomplished. In this case, the bus location is still reported, but its route is not. When this happens, NextBus simply discards the updates for that bus, and predictions do not take that bus into account. It is worth noting that we will be using the same data stream for our predictions as NextBus. 3. Proposed Work Upon completion, we hope to have evaluated two of the latest bus prediction algorithms and implemented one for the final application. To do this, we have acquired access to bus location update archives going back over 3 years, as well as source code for both the BuzzRoute and SMS bus tracker projects mentioned above. We have selected two bus-prediction algorithms to evaluate, one using a Kalman filter and the other a neural network. We have chosen to use MySQL as our spatial database solution and Java as our primary coding language. The rest of this section will explain these components in more detail. Bus Location Archives:Each bus, on the Georgia Tech transit system, is equipped with a GPS receiver and a cellular chip. Every 7 to 15 seconds the bus uploads its location to a server. This server combines the updates from all of the buses into a stream that NextBus accesses. In addition to being streamed to NextBus, the updates are also written to a daily log for archiving. As a result we have bus update records from as early as December of 2003. This is a sample update with an explanation of each line: 1. $GPRMC,050010.923,A,3346.6145,N,08423.6269,W,0.00,90.58,030107,,*2E // NMEA sentence2. I=1 // bus ignition status3. U=0 // deprecated and will not be used4. VID=810 // vehicle id5. RSSI=-71 // cellular signal strength6. NU=1:18:00 // time up on the network7. ND=0 // times rebooted the network card8. J=0202 // route id and block numberLine one is a NMEA sentence used by most GPS receivers. It is broken down as follows:RMC - Recommended Minimum Navigation Information 12 1 2 3 4 5 6 7 8 9 10 11| 13 | | | | | | | | | | | | |$GPRMC,hhmmss.ss,A,llll.ll,a,yyyyy.yy,a,x.x,x.x,xxxx,x.x,a,m,*hh<CR><LF>Field Number:


View Full Document

GT CS 4440 - Bus Prediction Algorithm Evaluation

Download Bus Prediction Algorithm Evaluation
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Bus Prediction Algorithm Evaluation and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Bus Prediction Algorithm Evaluation 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?