ORF467 Final Presentation Veenu King, Sylvia Okafor, Sally Yu TEAM XNJ
EXTERNAL NJ Because the focus of this assignment is NJ, each of our counties is represented by 1 pixel (very much like a centroid, or a station): Anyone who wants to take a trip must go to these pixels, and take a taxi/train from there. Is this realistic?
BUC NOR SOU WES NYC* INTL* PHL* *counties/pixels not shown
Train Trips
AVO
SOU
SOU our data is really weird… doesnt necessarily match up… probs because we are an external county and more data would be needed..?
BUC
BUC
WES
WES
NOR
NOR
Taxi Optimization Algorithm For each line of input data, make a trip, and give it it’s own personal taxi Compare the arrival taxis to departure taxis, and remove the departures that have a corresponding arrival Cycle remaining departure taxis to reduce fleet size before going into algorithm, describe data structures; result = rough estimate of how many taxis that would be needed for pixel (depending on existing demand in data)
Optimization Constraints Pixel -> pixel, CD = 3, DD = 300, Max circuity = 20% Assume an independent aTaxiService Company exists in each county
aTaxi Plan In order to meet demand, we will have garage(s)/lot(s) close to the station that holds the taxis not currently in use / waiting for deployment. Make sure there is at least one taxi of each size at station at all time, plus the taxis required by given demand distribution
SOU Optimization Results Total Departures from SOU: 6,455 Total Arrivals into SOU: 22,002 Foreign Taxis Used: 64 Min Number of Taxis Needed: 125
BUC Optimization Results Total Departures from BUC: 20,067 Total Arrivals into BUC: 70,597 Foreign Taxis Used: 726 Min Number of Taxis Needed: 790
WES Optimization Results Total Departures from WES: 3,079 Total Arrivals into WES: 10,745 Foreign Taxis Used: 113 Min Number of Taxis Needed: 29
NOR Optimization Results Total Departures from NOR: 2,719 Total Arrivals into NOR: 8,827 Foreign Taxis Used: 2 Min Number of Taxis Needed: 20
Room for Improvement... Unfortunately, because of the structure, there is a large empty mile burden, which could be solved, possibly, with more data (e.g. the taxi could check pixels/surrounding areas for a return trip, instead of coming back empty)
Room for Improvement... Updating the algorithm to be able to wrap over midnight when dealing with time Enhancing the algorithm’s calculation of taxi return time to be more accurate (which may reduce/increase the fleet size even more)
Room for Improvement... Make the algorithm more dynamic. Currently, algorithm gives a rough estimate of number of taxis needed based on the given demand in the data. However, anything can happen, so there should be (more) taxis to adjust for such cases. Problem is figuring out what said number is...