Professors: John Bartholdi (ISyE), Russ Clark (OIT, CoC), Bill Easton (OIT)
We will meet Tuesdays 10—11:30am in Groseclose 226 (the SCL/Georgia Freight Bureau seminar room in the ISyE building, on the same floor as academic advisement). If you cannot meet at that time you will have to withdraw from the project for now and try again next semester. (Special exception may be made for those with special skills and proven ability to work independently.)
Sub-project teams will meet as necessary and at their convenience.
There will be no exams. You will be judged by your professors and peers on your contribution to and engagement with the project (details).
Everyone is expected to maintain a lab notebook with a sewn or glued binding (example). They are available at Staples or the GT Barnes and Noble bookstore. The notebook will be graded as per these expectations.
During the first two weeks we will organize into groups to match interests and skills with goals and tasks for this semester.
This is a project course and so fundamentally different from most other courses. We have a client to whom we have made commitments; and progress depends on teamwork. Therefore the pace of the class is set by the needs of the project. Unlike most other classes, you cannot neglect your work and catch up later. The project does not stop during mid-terms.
Failure to deliver on commitments inconveniences the entire team and endangers the success of the project. Lapses that might seem incidental in a lecture class will not be tolerated here. Do not enroll for this class unless you are prepared to work towards a team goal.
The best way to ensure the success of the project and enjoy your work is to find a piece of the project in which you are intensely interested and then take it on. The object is to make a contribution that persists.
We encourage team members to return in following semesters, especially if they have shown potential to grow into positions of technical leadership.
The system has been running successfully for 3.5 years but we continue to improve it and to extend its capabilities.
GPS can be inaccurate, especially around tall buildings, which interfere with signals from the satellites. This makes it hard to note accurately when a bus arrives at a control point and – especially – when it departs.
Wendi Tang has built an app that will record all sensor data from an Android device and push it to a remote DB for analysis. It will also allow a user to record times of actual arrival and departure.
We will spend a lot of time this semester riding buses and using this app to record sensor readings plus actual arrival and departure times. Then we will analyze the collected data to see what combination of readings best correlates with arrival and departure.
Among the interesting sensor readings, besides GPS, are the accelerometer and the detection of Bluetooth beacons, which we will deploy.
The data collected here will also be used for the next task.
Can you do a better job than NextBus in predicting when the next bus will arrive?
Use dead-reckoning, machine-learning, or anything in between.
This will require the following steps:
Devise some clever way to estimate the number of passengers boarding or de-boarding at each stop.
If you have another idea you would like to tackle, make a proposal!
All our work lives on Github. To see it, you must log in with your user id and then email me your user id so I can grant access.
Look at the various projects underway and read their Wikis and examine their Issues. See any you want to work on? If not, propose something.
Do not start working without first having proposed what you will do, written a specification for it, and received approval.
If you are new to the project: