DSES-6620 Simulation Modeling Analysis Ernesto Gutierrez-Miravete Final Project: Traffic Light Hold-Ups An Adventure Down the Berlin Turnpike Submitted by: Jim Henriques December 2000 ABSTRACT The Berlin Turnpike consists of many traffic lights. Every morning people drive down it and have to stop at these traffic lights. A selected portion of the Berlin Turnpike is modeled in the simulation software package ProModel. Tests were performed in an attempt to find the speed that will result in the least amount of waiting time at red lights. Along with the overall time it takes to go through the system will help determine the optimal speed to travel during the morning commute on the Berlin Turnpike. 2 TABLE OF CONTENTS An Adventure Down the Berlin Turnpike .......................................................................... 1 1. Introduction ................................................................................................................. 4 1.1 Objective ................................................................................................................. 4 1.2 Scope ....................................................................................................................... 5 1.3 Requirements .......................................................................................................... 5 2. Implementation ........................................................................................................... 5 2.1 Data Collection ....................................................................................................... 6 2.1.1 Distance Collection ......................................................................................... 6 2.1.2 Traffic Light Data Collection.......................................................................... 6 2.2 Model Design .......................................................................................................... 7 2.2.1 Locations ......................................................................................................... 7 2.2.2 Entities ............................................................................................................ 7 2.2.3 Processing ....................................................................................................... 8 2.2.4 Arrivals ........................................................................................................... 8 2.3 Integration ............................................................................................................... 8 2.3.1 Assumptions.................................................................................................... 9 2.4 Testing the System .................................................................................................. 9 2.4.1 Test 1 ............................................................................................................... 9 2.4.2 Test 2 ............................................................................................................. 10 3. Results ....................................................................................................................... 10 3.1 Results: Test 1 ....................................................................................................... 11 3.2 Results: Test 2 ....................................................................................................... 11 3.3 Verification ........................................................................................................... 11 3.4 Validation .............................................................................................................. 12 4. Conclusion ................................................................................................................ 12 5. Appendix A: Distances Between Traffic Lights ....................................................... 14 6. AppenDix B: Traffic Light Times ............................................................................ 15 7. Appendix C: Test 1 Results ...................................................................................... 16 8. Appendix D: Test 2 Results ...................................................................................... 18 9. Appendix E: Sample Output and Trace file of System, Test 2 ................................. 19 3 1. INTRODUCTION Every morning countless people drive down the Berlin Turnpike in order to get to school, work, or just to run errands. These people all drive at different speeds, speeds that they are comfortable with. Some people are not in a hurry and drive at a slower pace while others wish to think that they are participating at the Indianapolis 500 and drive as fast as they possibly can. The one equalizing factor between these drivers is that they must obey traffic lights. Very often it seems that no matter what the speed of the driver is, two travelers going different speeds, through the timing of the traffic lights, end up at the same destination at relatively the same time. The Berlin Turnpike has been modeled to simulate the morning commuter traffic light patters to help gain insight into the optimal constant speed on should drive to get to their destination as safely and effectively as possible. This includes the time it takes to get there versus the amount of time one is waiting at a red light. 1.1 Objective The goal of this system is to determine the optimal speed to travel down the Berlin Turnpike during the morning rush hour commute. This includes taking into consideration how much time a commuter is actually waiting at a traffic light, and monitoring the percentage of time the vehicle is moving over the entire travel time. The optimal speed will be decided by the amount of time a vehicle is moving versus the amount of time it is stopped at a light. This will be correlated to the overall time of the commute. 4 1.2 Scope A portion of the Berlin Turnpike will be modeled in this system. The starting point to this system will be one mile away from the Turnpike. After this first mile, the Berlin Turnpike will be used for the commute. The portion of the Berlin Turnpike to be modeled will be from the first light in Berlin northward to the end of the Berlin Turnpike, where route 15 begins. It will consist in all of 27 traffic lights. 1.3 Requirements In order to portray an accurate representation of the Berlin Turnpike some pieces of information would be needed. The distances between each light would be necessary in order to model it correctly. The other main requirement is learning how the traffic lights work. Determining how long each traffic light on the Berlin Turnpike stays green, red and yellow. Also a working knowledge of a software package to simulate this system is necessary. In this case, the software package that will be used is ProModel 4.22. 2. IMPLEMENTATION In order to accomplish this system a plan needed to be developed. The plan identified what would be needed to fulfill the requirements. Once each step was identified, they were prioritized. The steps included collecting all pertinent data, developing a computer model, integrate the data collected into the system and then analyze the data from the model. 5 2.1 Data Collection There were two main pieces of data collection require for this system. The first was the distance between each traffic light and the second was the cycles of the traffic lights during rush hour. 2.1.1 Distance Collection The distances between the lights were recorded while driving down the Turnpike. Using a car odometer, the distances after going under each light were recorded. These distances are listed in Appendix A. 2.1.2 Traffic Light Data Collection The traffic lights cycles were obtained from the Connecticut Department of Transportation (DOT). Each light had its own blueprint and listed out how long they were green during the rush hour commute. With these blueprints it was determined how much time it was green, red and yellow on each side of the intersection. An employee at the DOT gave information on how to read them correctly. Each light has a different cycle depending on the time of day. Since this simulation was for a morning commute, the morning cycle was used (6:00 am – 9:00 am). That cycle then gives a number of how many seconds the light is green and the percentage of the time. There was also a section that showed how long it was yellow, and how long all of the lights were all red at the same time. In most cases the lights were yellow for 5 seconds, and the lights would all be red during the transitions for a period of 2 seconds. Although sometimes this data varied those are the average figures. All of the competing sides of the lights green times and the 6 2-second transition period were added up to give a total wait time at each light. This data would later be used as the downtime of the lights in the model. This data can be found in Appendix B. 2.2 Model Design Before using the data recorded a design of the system needed to be defined. Using ProModel test systems were developed. Once the concept was proven, integrating the recorded data to the system would be possible. 2.2.1 Locations It was decided that each traffic light would have to be a location as well as the stretch of road between each traffic light. This was done since a location is allowed to have a downtime. Each light would be given a downtime according to the amount of time it was scheduled to be red. The stretches between the lights were queues. This was done since they could be given a distance, and the speed was determined by the entity. A conveyor was also an option, but they have their own speed and the volume of them would make it difficult to test different speeds quickly. 2.2.2 Entities The only entity in this system was the vehicle that travels the system. A speed could be associated with the vehicle, which made it easy to change the speeds for testing optimal speeds. 7 2.2.3 Processing The processing was fairly simple in concept but tedious in implementing. The entity of the vehicle was to move from one location to the next in succession. It would go from Light_1 to Queue_1 and from Queue_1 to Light_2, and so on throughout the whole system. 2.2.4 Arrivals To test different times of the system a delay for arriving to the first location was used. Since the traffic lights cycles started at 6:00, a delay could be added to simulate someone driving through at any time. For instance by having a “First Time” of 90 minutes, it would simulate a ride down the Turnpike at 7:30. 2.3 Integration After the test system was developed with only two traffic lights and made up downtimes the real system could be created. Due to the high amount of locations, the Student Version of ProModel could not be used. Instead the Professional version was used. First the queues were created and given the correct attributes according to the data recorded for the distances between each light. The next part was to decipher the blueprints from the DOT. Each light location was then given a downtime of the calculated wait times from the collected data. A frequency was then set for how often it should be down. This time was the total green time plus the wait time. Once each light was finished the system was run. 8 2.3.1 Assumptions The assumptions made this is model were as follows: No other cars present on the road Braking and acceleration times could be averaged in constant speed Distance under traffic light were negligible Traffic light patterns occur as accurately as designed in blueprint Car can maintain a constant speed All speeds used are safe and within reason No traffic violations such as speeding or running a yellow light Yellow lights would be run up to one second before they turned red. 2.4 Testing the System With the system completed with all of the data collected a few different tests were run. The first test was to try different times of the morning. The second test was sending a vehicle through the system every minute for a two hour period, essentially between 6:008:00 am. Several different speeds were used, ranging from 35 mph, up to 70 mph, in different intervals. 2.4.1 Test 1 The first test used for different start times, 6:00, 6:30, 7:00 and 7:30 in the morning. Each time was used with twelve different speeds. The output of the system gave a total time of the vehicle in the system and the percentage that it was moving, and the percentage it was 9 stopped at a light. The table below shows a portion of the output. The full results can be found in Appendix C Time of Day 6:00 6:30 7:00 7:30 Fastest Time 70 mph (17.17 minutes/ 76.17%) 70 mph (20.31 minutes/ 64.41%) 70 mph (17.32 minutes/ 75.52%) 70 mph (17.27 minutes/ 75.73% Best Utilization 60 mph (17.25 minutes/ 84.36%) 35 mph (25.61 minutes/ 85.39%) 35 mph (25.65 minutes/ 85.28%) 35 mph (25.68 minutes/ 85.17%) Table 1. Best of Test 1 (minutes/utilization) 2.4.2 Test 2 The second test took an average of the morning rush hour commute. A vehicle was inserted into the system every minute over a two-hour time period. At the end of the simulation the average time and utilization was taken. Again, the same speeds were used as in the first test. As shown in Table 1, again, the fastest speed of 70 mph resulted in the fastest time, and the slowest speed, 35 mph, resulted in the best utilization. The full resul ts can be found in Appendix D. 3. RESULTS The results of the two tests show similar results. It appears that the faster one travels, the quicker one gets to the destination. However the utilization went down at the quickest time. The task then became finding the best utilization that correlate with the best speed. In many cases, the difference in time between 70 mph and 60 mph was only a few seconds, while the utilization difference was over 8% in favor of going the slower speed. Since the best utilization was normally 35 mph, it took ultimately too long to travel, so it was ruled out as optimal speed. At the other end of the spectrum, normally the utilization was not that great at the fastest speed, but some of the slower speeds only took a few 10 seconds less and their utilization’s were a few points higher and therefore they were selected to be the optimal speeds. 3.1 Results: Test 1 Table 2 below shows the best speed and utilization for the four times selected in Test 1. Time of Day 6:00 6:30 7:00 7:30 Speed 60 mph 60 mph 65 mph 63 mph Time in System 17.25 20.31 17.35 17.32 Utilization 84.36% 71.63% 79.3% 81.16% Table 2. Optimal Speeds From Test 1 From this data it appears that leaving at 6:00 in the morning and driving a constant speed of 60 mph would be the most effective. A sample output file for this system can be found in Appendix E. 3.2 Results: Test 2 The results for Test 2 show the average time in the system a vehicle going a constant speed would go. At any time during the rush hour commute it appears that going a constant speed of 63 mph would deliver the best speed and utilization combination of 18.95 minutes and 75.02% utilization. 3.3 Verification The system went through two verification periods. The first was during the initial design of the system. Once the test version performed as expected it was expanded into the final system. Once the final system was finished many test were run. A few minor defects 11 were found and quickly corrected. The system was setup to watch a car stop at a light with a known stop time. The car ran to the light and stopped for the appropriate amount of time and then continued. Since the system was of a basic setup repeated over and over again, since the small test worked, it is assumed the rest of the system will fall into place as the small test did. After watching the complete system run this was proven true. 3.4 Validation These number correspond fairly well with the actual amount of driving time it normally takes on this particular part of the Berlin Turnpike during the rush hour commute. Unfortunately it is not possible to perform this test in an actual car at the conditions as in the model. It has been attempted to go a constant speed but other traffic often is in the way thus ruining the experiment. Although it does seem like the system is producing logical travel times. 4. CONCLUSION The Berlin Turnpike was modeled well. It can be quickly calculated at any given point in the morning commute how fast one should drive for the optimal results. Unfortunately, there is some disappointment in the results that were gathered. It seemed that the faster one drove the quicker one got to their destination, regardless of the utilization. It was hoped that perhaps it could be proved that going slower could still take less time than going faster. While often it only a few seconds or a minute difference, it was always less time to go faster. Of course more can be looked into just the utilization of the system. For example over a long period of time a worse utilization may require a driver to use more 12 gas, and have more maintenance done to one’s car. If the cost of driving a car could be linked to the utilization number this experiment may still prove going slower 13 5. APPENDIX A: DISTANCES BETWEEN TRAFFIC LIGHTS Feet 5280 2640 4224 1056 792 2904 3696 2640 3168 1056 3168 1320 1848 1320 792 2112 528 2112 1320 792 1056 24.571 1848 264 5808 528 2112 Miles 1.00 0.50 0.80 0.20 0.15 0.55 0.70 0.50 0.60 0.20 0.60 0.25 0.35 0.25 0.15 0.40 0.10 0.40 0.25 0.15 0.20 0.00 0.35 0.05 1.10 0.10 0.40 Accumlated Miles 0.00 1.50 2.30 2.50 2.65 3.20 3.90 4.40 5.00 5.20 5.80 6.05 6.40 6.65 6.80 7.20 7.30 7.70 7.95 8.10 8.30 8.30 8.65 8.70 9.80 9.90 10.30 14 6. APPENDIX B: TRAFFIC LIGHT TIMES LightTotal # Cycle Time (SEC) Downtime (SEC) Uptime (SEC) 1 92 72 20 2 82 39 43 3 92 46 46 4 92 42 50 5 80 38 42 6 92 40 52 7 92 45 47 8 91 45 46 9 92 44 48 10 92 41 51 11 120 73 47 12 118 61 57 13 97 52 45 14 97 40 57 15 80 37 43 16 82 18 64 17 100 57 43 18 77 23 54 19 100 58 42 20 92 23 69 21 133.5 39.5 94 22 101 47 54 23 98 55 43 24 106 51 55 25 106 51 55 26 104 43 61 27 106 51 55 15 7. APPENDIX C: TEST 1 RESULTS Time: 6:00 MPH Total Time Total Time (Minutes) % Moving % Stopped 35 27.35 79.98 20.02 40 24.57 80.11 19.89 45 22.72 79.11 20.89 50 22.65 73.3 26.7 53 20.85 76.28 23.72 55 19.06 81.22 18.78 57 19.04 79.24 20.76 60 17.25 84.36 15.64 63 17.23 81.63 18.37 65 17.21 79.95 20.05 67 17.20 78.37 21.63 70 17.18 76.17 23.83 Time 6:30 MPH 35 40 45 50 53 55 57 60 63 65 67 70 Total Time (Minutes) % Moving % Stopped 25.62 85.39 14.61 25.62 76.82 23.18 25.62 70.15 29.85 23.85 69.61 30.39 23.85 66.68 33.32 23.85 64.92 35.08 23.85 63.27 36.73 20.32 71.63 28.37 20.32 69.21 30.79 20.32 67.73 32.27 20.32 66.34 33.66 20.32 64.41 35.59 16 Time Time 7:00 MPH 35 40 45 50 53 55 57 60 63 65 67 70 7:30 MPH 35 40 45 50 53 55 57 60 63 65 67 70 Total Time (Minutes) 25.65 24.63 24.55 22.72 22.67 22.66 22.64 19.12 19.10 17.35 17.34 17.33 Total Time (Minutes) 25.68 24.67 22.82 22.75 20.95 20.93 20.91 20.88 17.33 17.31 17.30 17.28 % Moving 85.28 79.89 73.2 73.09 70.12 68.32 66.64 76.12 73.63 79.3 77.72 75.52 % Moving 85.17 79.78 78.76 72.98 75.92 73.98 72.17 69.68 81.16 79.49 77.92 75.73 17 % Stopped 14.72 20.11 26.8 26.91 29.88 31.68 33.36 23.88 26.37 20.7 22.28 24.48 % Stopped 14.83 20.22 21.24 27.02 24.08 26.02 27.83 30.32 18.84 20.51 22.08 24.27 8. APPENDIX D: TEST 2 RESULTS Average Time Throughout 2-hour Period Speed Time (sec) Time (min) Moving % 35 26.82 73.68 40 25.18 81.96 45 24.16 78.55 50 22.34 74.82 53 21.21 74.9 55 20.44 75.57 57 20.08 76.39 60 19.57 75.78 63 18.96 75.02 65 18.72 74.79 67 18.51 74.1 70 17.95 73.5 Stopped % 26.32 18.04 21.45 25.18 25.1 24.43 23.61 24.22 24.98 25.21 25.9 26.5 18 9. APPENDIX E: SAMPLE OUTPUT AND TRACE FILE OF SYSTEM, TEST 2 19