1 00:00:08,500 --> 00:00:12,600 Hi, I am Alexander Verbraeck, professor of systems and simulation and I'm 2 00:00:12,619 --> 00:00:19,000 going to look with you to a number of discrete event simulation examples following up on 3 00:00:19,000 --> 00:00:24,050 the discrete event simulation module that you have looked at before. 4 00:00:24,050 --> 00:00:29,119 In this module I'm going to look at a number of differences between discrete event simulation 5 00:00:29,119 --> 00:00:33,749 software types and I'm going to look at three examples with you: 6 00:00:33,749 --> 00:00:38,300 one on an airport, to look at logistics simulation one on a large 7 00:00:38,300 --> 00:00:41,620 international supply chain and one on barge transportation. 8 00:00:41,620 --> 00:00:47,510 I will end with some conclusions and an outlook for discrete event simulation. 9 00:00:47,510 --> 00:00:51,039 If we look at the types of simulation software available, 10 00:00:51,039 --> 00:00:53,479 there are actually two different types. 11 00:00:53,479 --> 00:01:00,010 One is programming languages that we can use to build simulation software. 12 00:01:00,010 --> 00:01:02,920 Libraries are used to make the simulation software, 13 00:01:02,920 --> 00:01:05,969 but they're not so easy to use for end-users. 14 00:01:05,969 --> 00:01:10,890 Therefore, there are also a large number of general purpose simulation packages available 15 00:01:10,890 --> 00:01:12,549 on the market. 16 00:01:12,549 --> 00:01:17,869 Examples are Arena, Simio, Simulate, Plant Simulation, Enterprise Dynamics, 17 00:01:17,869 --> 00:01:20,180 Extend and AnyLogic, and there are many, many others, 18 00:01:20,180 --> 00:01:23,950 there are hundreds of different types of simulation software on the market. 19 00:01:23,950 --> 00:01:28,200 Some of them are more suitable for some types of infrastructure models and others more suitable 20 00:01:28,200 --> 00:01:30,859 for other types of infrastructure models. 21 00:01:30,859 --> 00:01:33,640 If we look at the developments in the simulation packages, 22 00:01:33,640 --> 00:01:37,369 we see quite a number of fascinating things happening. 23 00:01:37,369 --> 00:01:38,770 One is the use of libraries. 24 00:01:38,770 --> 00:01:45,770 We already covered that a little bit in the lecture on discrete event simulation modeling. 25 00:01:46,229 --> 00:01:47,829 Hierarchy is another one. 26 00:01:47,829 --> 00:01:51,009 Building models that can become more and more complicated. 27 00:01:51,009 --> 00:01:58,009 3D animation is more and more the norm today for discrete event simulation packages. 28 00:01:58,670 --> 00:02:02,999 Furthermore, all kinds of input and output from and into databases, 29 00:02:02,999 --> 00:02:09,500 excel, statistical packages and all kinds of other input and output software Link to 30 00:02:09,500 --> 00:02:13,250 optimization software and all kinds of extensions. 31 00:02:13,250 --> 00:02:19,940 Lately a number of packages - among which AnyLogic - also offers multi-formalism modeling. 32 00:02:19,940 --> 00:02:23,000 That means that we can mix discrete event simulation models, 33 00:02:23,000 --> 00:02:28,000 continuous simulation models and agent-based simulation models for the behavior of people 34 00:02:28,000 --> 00:02:30,550 in our models. 35 00:02:30,550 --> 00:02:32,570 Let's look at our first case. 36 00:02:32,570 --> 00:02:34,360 A case for airport logistics. 37 00:02:34,360 --> 00:02:41,350 What we observed in a number of studies is that every time we had to build a model for 38 00:02:41,350 --> 00:02:45,150 solving problems at an airport we had to make a new model. 39 00:02:45,150 --> 00:02:47,180 We had to code it and we had to get the data. 40 00:02:47,180 --> 00:02:52,260 This means that it was a long process every time to build a model. 41 00:02:52,260 --> 00:02:54,480 Our challenge in this project was: 42 00:02:54,480 --> 00:02:58,560 can we construct a model out of building blocks? 43 00:02:58,560 --> 00:03:02,430 Small blocks from which we can build a model bottom-up? 44 00:03:02,430 --> 00:03:07,950 Can we thereby tackle all kinds of airport problems in a much more generic way? 45 00:03:07,950 --> 00:03:12,380 Our goal was to develop one set of simulation libraries for airport logistics, 46 00:03:12,380 --> 00:03:16,320 for airport design and for airport development. 47 00:03:16,320 --> 00:03:22,320 In the end, we created a set of library building blocks from which you can see here the highest level. 48 00:03:23,300 --> 00:03:28,110 So you see different concourses appears, different departure holes and at the bottom 49 00:03:28,110 --> 00:03:33,950 of the screen you can actually see all kinds of control blocks that control how people 50 00:03:33,950 --> 00:03:39,570 move between different parts of the airport, how the planes behave, 51 00:03:39,570 --> 00:03:43,610 when they land, what their flight schedule is, etc. 52 00:03:43,610 --> 00:03:48,900 When we zoom in on one of these building blocks, for instance on one of the concourses, 53 00:03:48,900 --> 00:03:50,590 we see a smaller model; 54 00:03:50,590 --> 00:03:54,260 a couple of gates, a couple of walk areas, 55 00:03:54,260 --> 00:04:01,260 a couple of conveyor belts that help the passenger to go from or to the gate and from there to 56 00:04:02,070 --> 00:04:03,780 other parts of the airport. 57 00:04:03,780 --> 00:04:09,830 So this is zooming in into one of the building blocks that you saw on the previous slide. 58 00:04:09,830 --> 00:04:13,370 If we zoom in even more, for instance on one of the gates, 59 00:04:13,370 --> 00:04:20,370 we see that the gate consists of a wait area, that means that people wait to be checked. 60 00:04:20,810 --> 00:04:25,820 Then they wait at the gate to get their boarding passes checked an through the bridge they 61 00:04:25,820 --> 00:04:29,260 can actually leave the model into the plane. 62 00:04:29,260 --> 00:04:33,900 Each of the building blocks that we saw on this particular slide, 63 00:04:33,900 --> 00:04:39,840 each of the gates, F1, F2, F3 etcetera, consists of a building block like this. 64 00:04:39,840 --> 00:04:44,370 These building blocks are very similar and are re-used in the model. 65 00:04:44,370 --> 00:04:48,040 The one thing that we did for this particular model is that we can parameterize the building blocks. 66 00:04:48,650 --> 00:04:53,000 It means that we can set all kinds of properties of the building block to make them behave 67 00:04:53,000 --> 00:04:57,800 exactly the way it should be for this particular gate. 68 00:04:57,800 --> 00:05:02,340 One of the interesting things is that we can also animate the model on the highest level. 69 00:05:02,340 --> 00:05:08,220 It means we don't see all the individual gates and all the individual building blocks on 70 00:05:08,220 --> 00:05:12,710 the animation, but we actually see the concourse the way 71 00:05:12,710 --> 00:05:14,810 that anybody would see it. 72 00:05:14,810 --> 00:05:18,500 We see that people move towards the gates, we see them walking, 73 00:05:18,500 --> 00:05:25,500 we see them using the conveyor belts and this means that we have a very, 74 00:05:25,650 --> 00:05:30,700 very good indication of how people move, how people walk and what they do. 75 00:05:30,700 --> 00:05:35,120 We see it actually getting more and more busy - also at the wait areas - and it's only a 76 00:05:35,120 --> 00:05:42,000 matter of time until the first planes depart and we have actually people going out of the model. 77 00:05:42,600 --> 00:05:47,900 One of the most important things is of course the fact that we can also produce output from the model. 78 00:05:48,110 --> 00:05:53,000 Here we actually see the model that we've just watched in terms of the number of passengers 79 00:05:53,000 --> 00:05:56,280 at a certain desk row for checking in. 80 00:05:56,280 --> 00:05:59,850 We can also look at all kinds of other statistics. 81 00:05:59,850 --> 00:06:03,010 Statistics are one of the main reasons for creating simulation models, 82 00:06:03,010 --> 00:06:05,710 and especially discrete event simulation models. 83 00:06:05,710 --> 00:06:10,580 When we look back at this particular case, we see that we created one library for the 84 00:06:10,580 --> 00:06:12,180 passenger terminal logistics. 85 00:06:12,180 --> 00:06:16,800 We can very quickly model the infrastructure due to the fact that we have high-level building blocks. 86 00:06:17,050 --> 00:06:23,900 The hierarchy helps is to reuse earlier efforts and to make sure that we don't have to invest 87 00:06:23,960 --> 00:06:28,389 time again in rebuilding building blocks that we made before. 88 00:06:28,389 --> 00:06:32,060 The models are still complex though and a lot of behavior is hidden. 89 00:06:32,060 --> 00:06:36,460 It still is difficult for people to create a good model from this. 90 00:06:36,460 --> 00:06:38,520 They need training. 91 00:06:38,520 --> 00:06:41,490 One of the other things that needs more focus is: 92 00:06:41,490 --> 00:06:44,060 input, output and scenarios. 93 00:06:44,060 --> 00:06:47,090 And that's what I will show in the next case study. 94 00:06:47,090 --> 00:06:51,840 The next case study is about global supply chain management. 95 00:06:51,840 --> 00:06:54,710 Global supply change management is extremely complex. 96 00:06:54,710 --> 00:07:00,040 Goods coming from China being imported to Europe with all kinds of raw materials from 97 00:07:00,040 --> 00:07:04,010 all over the world, that are then again exported from Europe across 98 00:07:04,010 --> 00:07:05,220 the globe. 99 00:07:05,220 --> 00:07:06,900 Global supply chains. 100 00:07:06,900 --> 00:07:10,060 Many changes, with time as a very important factor, 101 00:07:10,060 --> 00:07:15,889 and extremely complex business relationships including competition. 102 00:07:15,889 --> 00:07:20,550 We really want to compare alternatives in all kinds of different scenarios and the question is: 103 00:07:20,830 --> 00:07:22,830 how can you parameterize those scenarios? 104 00:07:22,830 --> 00:07:29,400 What you see on the screen here is a very simple model at the top in terms of the weights represented. 105 00:07:29,520 --> 00:07:34,240 The model at the bottom, however, shows the background of the particular model, 106 00:07:34,240 --> 00:07:38,520 the way it is being built, and we can see that it is extremely complicated. 107 00:07:38,520 --> 00:07:42,449 The question is: How can we make this more simple for the users. 108 00:07:42,449 --> 00:07:46,639 How can we create a flexible set of models for demonstration and teaching of these very, 109 00:07:46,639 --> 00:07:51,040 very complicated global infrastructures? 110 00:07:51,040 --> 00:07:54,930 This is a view of the way a model looks: 111 00:07:54,930 --> 00:07:58,830 We see the fact that on the left side we have initial suppliers, 112 00:07:58,830 --> 00:08:03,169 we have on the right hand side all consumers and customers and in the middle the focal 113 00:08:03,169 --> 00:08:05,120 company we want to look at. 114 00:08:05,120 --> 00:08:10,610 And this focal company we want to learn more about in a lot of our studies. 115 00:08:10,610 --> 00:08:15,330 This is an example of a teaching case that we built based on the library again of all 116 00:08:15,330 --> 00:08:19,760 kinds of building blocks in which we compared two different strategies. 117 00:08:19,760 --> 00:08:24,110 On the top, I made the stock supply chain and on the bottom I made the order supply chain. 118 00:08:24,520 --> 00:08:28,510 The difference between these supply chains is whether we pull or we push the orders. 119 00:08:28,510 --> 00:08:33,099 Are the orders pushed from the left hand side, you actually push your product to the markets. 120 00:08:33,099 --> 00:08:36,189 Are the orders pulled from the left hand side means, 121 00:08:36,189 --> 00:08:40,019 that you actually work in the other direction. 122 00:08:40,019 --> 00:08:44,240 It leads to a very different way of sending information, 123 00:08:44,240 --> 00:08:49,290 receiving information and sending for instance products. 124 00:08:49,290 --> 00:08:54,319 Again in this particular example, the output is of the utmost importance. 125 00:08:54,319 --> 00:08:57,899 We see a lot of different graphs that we can create with this particular model and they 126 00:08:57,899 --> 00:09:03,119 provide a lot of insight into what happens in the different parts of the supply chain. 127 00:09:03,119 --> 00:09:08,249 We can look at inventory positions, we can zoom-in to individual companies and 128 00:09:08,249 --> 00:09:13,869 we can look at all kinds of information that are important to test the effect of the decisions 129 00:09:13,869 --> 00:09:18,749 we've just made. 130 00:09:18,749 --> 00:09:21,920 We extended these particular models to gaming. 131 00:09:21,920 --> 00:09:26,449 It means that the same models that you saw the previous slides are at the core of a number 132 00:09:26,449 --> 00:09:30,670 of supply chain games that we created to play with students. 133 00:09:30,670 --> 00:09:34,509 This is an example of a screen of a global supply chain game that we actually played 134 00:09:34,509 --> 00:09:40,119 globally with students and in the material on the MOOC you can actually find more information 135 00:09:40,119 --> 00:09:42,439 about this. 136 00:09:42,439 --> 00:09:48,990 Students play a relatively simple business case in many different parts of the world. 137 00:09:48,990 --> 00:09:53,449 And underneath are exactly the same building blocks of discrete event simulation that you 138 00:09:53,449 --> 00:09:56,769 saw on the previous slides. 139 00:09:56,769 --> 00:10:00,319 With this, we provided a flexible solution for supply chain management. 140 00:10:00,319 --> 00:10:04,850 The side by side comparison that you saw in the animation provides all kinds of insight 141 00:10:04,850 --> 00:10:06,790 for the users. 142 00:10:06,790 --> 00:10:10,920 We focused very much on the output here, contrary to the models that you saw in the 143 00:10:10,920 --> 00:10:12,689 first case study. 144 00:10:12,689 --> 00:10:18,009 In this particular case we used Java libraries to build the models to have full flexibility 145 00:10:18,009 --> 00:10:20,869 in the way we could build them. 146 00:10:20,869 --> 00:10:27,139 Finally, serious games have been developed with discrete simulation models as the core, 147 00:10:27,139 --> 00:10:32,139 which aligns very much with decision making as human decision makers do and they can enter 148 00:10:32,139 --> 00:10:35,350 their decisions into the game. 149 00:10:35,350 --> 00:10:38,339 Our final case study is about a barge transportation. 150 00:10:38,339 --> 00:10:41,589 In this particular case our challenge was: 151 00:10:41,589 --> 00:10:47,149 can we create micro simulation models with a lot of detail that we can use to support 152 00:10:47,149 --> 00:10:51,069 decisions 10, 15 or 20 years ahead. 153 00:10:51,069 --> 00:10:53,309 Long term policy studies. 154 00:10:53,309 --> 00:10:56,559 The question was: are these models fast enough and would they 155 00:10:56,559 --> 00:10:59,339 be usable in this particular setting? 156 00:10:59,339 --> 00:11:03,579 In this particular case we created the micro level simulation model for barge transportation 157 00:11:03,579 --> 00:11:10,579 in the Netherlands usable in policy setting that can be used 10 to 20 years ahead. 158 00:11:10,639 --> 00:11:15,829 The model itself is based on a geographical information system and uses a lot of different 159 00:11:15,829 --> 00:11:20,869 databases to support the model background. 160 00:11:20,869 --> 00:11:27,819 The data itself - barge transportation, terminals, waterways, volumes - 161 00:11:27,819 --> 00:11:33,990 are based on all kinds of measurement and statistical data and also on official records. 162 00:11:33,990 --> 00:11:36,240 When we slow down the simulation time a little bit, 163 00:11:36,240 --> 00:11:40,029 you can actually see individual ships sailing and the moment I click on one, 164 00:11:40,029 --> 00:11:44,600 you can see the information about the ship, the number of containers it contains, 165 00:11:44,600 --> 00:11:46,699 where it's coming from and where it's going to. 166 00:11:46,699 --> 00:11:50,249 Each individual containers is modelled for this particular simulation mode, 167 00:11:50,249 --> 00:11:52,189 several millions a year! 168 00:11:52,189 --> 00:11:56,999 And still the model actually completes a year of simulation in about one minute. 169 00:11:56,999 --> 00:11:59,160 One of the advantages of discrete event simulation. 170 00:11:59,160 --> 00:12:02,970 We can also see information about the terminals. 171 00:12:02,970 --> 00:12:07,749 This means that we can look at the stack positions in the terminals, 172 00:12:07,749 --> 00:12:11,709 how many containers are stored, how many key cranes and other types of cranes 173 00:12:11,709 --> 00:12:15,959 do they have and we can look at all kinds of other information. 174 00:12:15,959 --> 00:12:18,869 One of that is the locks and bridges; 175 00:12:18,869 --> 00:12:24,839 I just clicked on one of the pointers on the left hand side to turn on the information 176 00:12:24,839 --> 00:12:26,699 on bridges and locks. 177 00:12:26,699 --> 00:12:31,889 The locks and bridges can also be questioned and queried and we can see for instance whether 178 00:12:31,889 --> 00:12:36,809 certain types of ships would be able to sail those particular locks and bridges. 179 00:12:36,809 --> 00:12:40,809 This helps us to take all kinds of infrastructure decisions for the future, 180 00:12:40,809 --> 00:12:47,199 where we can for instance look at the effect of investing in extension of locks, 181 00:12:47,199 --> 00:12:51,499 or making bridges higher so we can actually sail with larger ships. 182 00:12:51,499 --> 00:12:57,529 We can also look at the effect of widening or deepening certain waterways or canals. 183 00:12:57,529 --> 00:13:04,000 And all these things can be done with one simulation model that we used in quite interactive settings. 184 00:13:05,000 --> 00:13:08,199 On the right hand slide, you see one of the settings that we did. 185 00:13:08,199 --> 00:13:13,649 A setting with Rijkswaterstaat in the Netherlands (part of the ministry of Infrastructure and 186 00:13:13,649 --> 00:13:18,279 Environment) where we looked at the use of these micro simulations in decision making 187 00:13:18,279 --> 00:13:23,290 settings and where we studied the effect of all kinds of infrastructure changes towards 188 00:13:23,290 --> 00:13:24,550 the future. 189 00:13:24,550 --> 00:13:29,319 We really were able to support long term decision making with these particular models. 190 00:13:29,319 --> 00:13:34,769 This discrete event simulation formalism took care of creating very fast models. 191 00:13:34,769 --> 00:13:38,660 Participants could use the models and were fully engaged in the sessions (by the way 192 00:13:38,660 --> 00:13:43,369 the sessions that we did lasted one to two days in which all kinds of infrastructure 193 00:13:43,369 --> 00:13:46,110 decisions were tested). 194 00:13:46,110 --> 00:13:51,600 Thank you very much for your attention and I hope that you will learn a lot more about 195 00:13:51,600 --> 00:13:55,999 discrete event simulation using the materials that are available on this MOOC. 196 00:13:55,999 --> 00:13:58,600 Thank you very much.