Emily Rogers Jean Wang ORF 467 Final Report-Middlesex County Abstract The purpose of this investigation is to model the demand for an ataxi system in Middlesex County. Given transportation statistics for a typical October weekday the effectiveness and potential constraints on an ataxi system were investigated. The analysis included demand by vehicle type, with wait time and common destinations constrained, the number of vehicles potentially required, and the number of empty vehicle miles for all vehicles. Executive Summary Middlesex County has a population of 828,919 1 people and is home to Rutgers University. As a result 20% of the population is between the ages of 18 and 24. This leads to some interesting patterns in transportation demand. College students tend to commute less in the morning, and stay out later at night. The number of vehicle moving with passengers actually peaks at 8pm. Additionally; only 10% of the population is below the age of 18, which is likely why there is a decrease in traffic around 3pm, instead of an increase seen in more suburban areas. There are just under three million person trips originating in Middlesex every day. At the peak of travel there are about 75,000 vehicles moving with passengers in them. A majority of them are six passenger vehicles, which are by far the most dominant vehicle type. There are over 40,000 six-passenger vehicles on the road for six hours a day. Comparatively two-passenger vehicles are used less (a peak of 14,000) but are fully utilized for most of the day. 20 and 50 passenger vehicles on the other hand have a single peak, and then are mostly idle for the rest of the day. This potentially means that eliminating large passenger vehicles in favor of two and six passenger ones may be more cost effective. To find ceiling numbers for the required vehicles for an ataxi system. We assumed vehicle relocation happened only at night, and that vehicles did not return from outside the county until the end of the day. If a pixel did not have a vehicle when a passenger arrived, one was drawn from one of the eight surrounding pixels. There was no extreme effort to reduce number of vehicles, or empty vehicle miles. 1 Data comes from the 2010 Census
2 passenger: 272,475 6 passenger: 483,530 20 passenger: 15,824 50 passenger: 846 2 and 6 passenger vehicles use the most number of vehicles compared to trips made. They also have the most trips per day. Constant relocation could significantly reduce the required number of vehicles by almost an entire order of magnitude. Travel Density By Time and Type To begin allocating vehicles we started by looking at how the demand for vehicles changes throughout the day. Summing the total number ataxis travelling at a given time was then graphed against hours from midnight. Only ataxis with passengers in them were counted. This shows ataxis of all types moving throughout the day. Rutgers University contributes a large part of Middlesex County s population. This is likely why the morning commute is significantly less than the peak at 8:30pm. Students commute later and are more likely to leave for afternoon classes. They also contribute to the lateness of the evening commute. The youth of the county is also likely why there is actually a drop in travel around 3:00pm when students would be getting out. Fewer residents have children. College students also contribute to the high volume of ataxis on the road after midnight.
Two passenger vehicles are by far the most flexible of the ataxis. The demand is by far the most consistent and is clearly preferred for late night travel. The fairly consistent usage throughout the day indicates that they are a more efficient vehicle and would most benefit from empty vehicle management. The number used is fairly low compared to 6 passenger vehicles however. This could mean they could help take over more trips during the peak hours for 6 passenger vehicles, to reduce the total number of ataxis needed Six passenger vehicles do have a significant drop in usage, but it is during a fairly short period of time and usage is rather consistent during business hours. Repositioning could be done during those lulls, and may require minimal empty vehicle management other than that.
Both 20 and 50 passenger vehicles follow similar trends in usage. Both are heavily used for the morning commute and then average around 10-20% max usage for the rest of the day. The discrepancy in usage between morning and evening commutes is the main source of inefficiency. Since people do not always return straight home from work high occupancy vehicles sit idle after 9am. 50 passenger vehicles go completely unused for a majority of the day. Based on this, and depending on the cost, it may make sense to simply not use high occupancy vehicles. They would rack up a significant number of empty vehicle miles, since there are few return trips and would need significant repositioning at the end of the day. This indicates that maybe 6 passenger vehicles should take over the morning commute since they are only at about half their maximum usage from 7-9am.
Single Pixel Investigation The most active pixel in Middlesex County is home to Rutgers University and has over five times as much activity as the next pixel. Relatively few people live in the pixel, so most trips are to and from work. Additionally there is a large train station that serves the Northeast Corridor line. People use to travel to New York and Philadelphia Cumulative Arrivals and Departures for 2 passenger ataxis
There are virtually no 2-passenger vehicle departures from the pixel 2. This is likely due to the train station, and the lack of nightlife in the pixel. Cumulative Arrivals and Departures for 6 passenger vehicles There are virtually no departures from the pixel until the end of the workday. Arrivals begin with the morning commute around 8am. There are more departures than arrivals, which might be from people arriving by train then leaving in smaller vehicles. Upper Limit Vehicle Requirements by type The upper limits for types of vehicles needed are the number of vehicles need, assuming no empty vehicle management except at the end of the day. Vehicles that leave the county stay there until the end of the day. To find a vehicle is the current pixel does not have any we looked at the surrounding pixels and calculated the time for them to arrive. Travel time is assumed to be 1.2*distance*2 minutes per mile. For adjacent pixels travel time is 1.2*1*2 = 2.4 minutes, and for diagonal pixels travel time is 1.2*2^.5*2 = 3.39 minutes. A vehicle could come from slightly farther away, which would reduce the need for repositioning. The decision from where to pull the car does become more complicated however. 2 Just a note, we triple checked this number. It is right according to the data we were given
2 passenger: 272,475 6 passenger: 483,530 20 passenger: 15,824 50 passenger: 846 Conclusion and Possible Improvements The first major area for improvement would be returning vehicles from outside the county before the end of the day. The simplest way would be to have the vehicle return to its origin pixel after waiting for ten minutes. A better solution would be to have it return to a pixel with high demand. This demand could be calculated a couple of ways. Vehicle could go to pixels with the highest population, or they could go to pixels with the greatest difference between their start number of vehicles and current vehicles. Either of these would be fairly simple to implement. The best way would be to have the car go to a pixel with more departures than arrivals in the coming hours. This would require looking at historical data, and looking into the future, but would put cars in the best positions for further use. After returning vehicles from other counties, empty vehicle management within the county would most help reduce the number of vehicles needed. There are several ways it could be approached, but one of the big deciding factors would be cost. If vehicles are relatively cheap to buy, but costly to run it would make more sense to purchase lots of vehicles and not reposition them until the end of the day. This would reduce the number of empty vehicle miles, but would require a lot of vehicles. In the more likely scenario that vehicles are expensive and running them is a significantly more minor expense repositioning throughout the day would make the most economic sense. The methods that could be employed are similar to those used for returning a vehicle from outside the pixel. The most efficient would be to have some vehicles that arrive in low departure pixels go to the nearest pixel with an arrivals deficit. This would require looking at historical data, and selecting how many and which vehicles to reposition, and when would not be easy to code. A simpler way may be to have vehicles that arrive in a pixel that has a surplus of vehicles (determined by some function) go to the nearest pixel with a deficit. Empty vehicle management is the best way to reduce the cost of an ataxi system in Middlesex County. Another way to reduce the costs would be to examine each type of vehicle. For example 50 passenger vehicles sit idle most of the time. If they were eliminated could those trips be transferred to 2, 6, and 20 passenger vehicles effectively? It is quite possible that the expense of large occupancy vehicles is larger than that of buying more, smaller vehicles to pick up the slack. This would all depend on the cost to buy and maintain each type of vehicle. The idleness of each type would also have
an impact on the effectiveness of this method of management. Possibly the best way to reduce the costs of an ataxi system is to run a whole lot of simulations, and pick the cheapest one based on real world costs.
Appendix 1 Vehicle Distribution at end of day in all of New Jersey 2 Passenger Vehicle End of Day Difference 6 Passenger Vehicle End of Day Difference 20 Passenger Vehicle End of Day Difference 50 Passenger Vehicle End of Day Difference