[00:00:05]
>> We are it's you know. Our series. It's my pleasure today to introduce one no from our latest edition to the I it's quite. A skull found him 33. Russell Chandler the 3rd chair professor I asked why he and him were always 3 weeks that he joined this.

[00:00:34]
For those who few who don't know he yet he brings a strong computer science background in addition to being a stroll along person and that's clearly very valuable Here's a way that we provide but also in the research that we do. Do you inform to use a fellow off Hey I already have a society.

[00:01:03]
Work. Well the strain. Lately. All. But thank you thank you Martin thank you for coming. Some going to stay at a very high level today but you can interrupt me any time for questions. Don't be shy so this is what I'm going to talk about I'm going to talk about the challenges in and the opportunities in mobility and then I'm going to give you 3 examples one which is this on demand transit systems that we are building another one is going to be community based trip sharing and then I'll talk a little bit of both real time Dyer to ride problems which is essentially what Wilbur and lift are doing and I really have no time to talk about constrained by a different kind of privacy but you know you can come to me if that's of interest to you so what are the challenges you know ports and opportunities in in mobility so one of the things that you have is that congestion I mean we do not plan to write so you know congestion is must if you're just today I took about 17 minutes to drive 6 miles not very good and this is this is supposed to be increasing and can go up to $200000000000.00 in a in the next 15 years it affects large and small cities and I'm sure you recognize one of them today.

[00:02:31]
So the other things that he'll have his dad at this point about 30 percent of the greenhouse gas emission in this country is due to transportation. And all of these 30 percent about 60 or 66 percent to 30 of them come from small vehicles and trucks so if we get if we remove these greenhouse gas emission we essentially addressing a very soon substantial way climate change.

[00:02:58]
Some cities Iraq to meet changing mobility just because of that reason this is an article a couple of a couple of months ago in Europe in Paris saying that they were considering free public transportation for everyone to just reduce the pollution in Paris. One of the big challenges in a big challenge in mobility is what is called the 1st and the last my problem so what you see on this light is the market share of transit system in terms of the distance that people have to walk to the bus stop OK And so what you can see is if you make people walk about a quarter of a mile you lose a century 90 percent of your ridership except for some reason in Calgary in Canada it's actually very cold there but they are willing to walk much more than anybody else don't ask me what I have to do I know why but I can tell you so what this child so shows you is that you have to essentially make people work less than 50 meters to a transit system otherwise they will not use it again so that you have to be very very close.

[00:04:02]
So why is this important it's important because essentially if you have mobility you have access to a lot of different things and if you don't you don't have access to those things so mobility you know transportation is actually the best predictor of social poor social mobility in the United States if you have a car you're much better off that if you don't there are millions of people also with health insurance and they cannot go to the K.-Y. because they don't have the mobility options and I can continue like that for many things they are 30000000 people I think 20 Pretty 3000000 people Internet States we don't leave within one mile of a supermarket So what do they do this shopping convenience store so they have you know bad quality food there and therefore they use and then they go back to the previous problems that I've that I've talked about so all these things.

[00:04:48]
Basically you know this a basic need of fundamental things that you suffer from when you don't have mobility This is actually a really significant issue in some cities so this is a this is actually a supermarket in Detroit and they were essentially offering overwrites for everybody spending more than 50 bucks inside a supermarket so can you imagine the mobility issue that these people had so fortunately we live in a very interesting world and there are a number of technology technology a neighbor making a huge difference and we're going to be surprised by the 1st one so I think the most important thing in mobility these days is this so it's communication so at this point essentially what you have he's a very interesting situation where we know where people are we know the infrastructure what the infrastructure is and what is the infrastructure is doing and what every very close around the city are doing and this is very very powerful as you'll see.

[00:05:46]
The 2nd things that are going to come at some point is does automated vehicles and they're going to make a difference and I will quantify doubt at some point in this talk but at some point they will come in one form or no on another and they will make a very significant difference not only in safety but also in terms of the cost that.

[00:06:02]
The mobility system we incur And finally and this is what we're doing in this department of your city school. We have made a huge amount of progress in of the musician in machine learning in the last in the last 2 decades so what we can do no we've the sole verse is amazing compared to what we could do 10 or 15 years ago another Lois to tackle problems that are much much larger in scope.

[00:06:26]
So and and that's obviously what we are mostly doing so let me show you a couple of examples snow and the 1st one is this on the mantle to Mother public transit and so one of the way to fink about what we're doing there is organizing a supply chain but instead of a supply and shames of goods it's a supply chain of people OK you know people a little bit more complicated pockets and I'll talk a little bit about that but you know fink about this is we are going to using a big supply chain of people so the system that I'm going to show you is that is on demand so that we under the 1st and the last my problem OK so we're going to pick up people where they live very close to where they live OK I'll show you what what I mean by that and then is going to be multimodal for so that we have a congestion so in a high density coal resource we don't run small car so small trucks we run high capacity vehicles such that we don't we avoid creating as much congestion as as we can and then to avoid greenhouse gas emission we want the system to be completely electric fight and you see that these systems actually remove the main the main difficulty in electrification which is wrong here we control the fleet so we don't have any ranches ranch and society.

[00:07:37]
And finally did raise the cost aspect and the cost that we want for this system is exactly the cost of a transit system so $3.00 or 2 dorsal $4.00 or something like that so in order of money to cheaper not only our game so how do we do this so I told you that this is like a supply chain of people so we have these small shutdowns here that are basically picking up people with their life and bringing them to their destinations already if they have to go through a very dense corridor then they will go into a high capacity vacant stand then they will go to the next up we're assured that he's waiting for them and we bring them to their destination that's the model now this is 2 levels but we can have 3 levels if we have a rail you know this is typically what you would have about planned if you have the buses and the rail and the shuttles OK Does the model know a couple of things that these to be done is one ticket online space an integrated system you buy your ticket on you you know you or you buy your ticket online you use your phone to actually abided by that by your right and then you don't have to worry about anything else the price as I told you is the price of a transit system and it works you'll see a financially and a transfer a completely synchronizes not like step of the bus and you say hey where do I go you will know your phone will tell you essentially what is the next shuttle that you have to do or what which is the bus you have to book.

[00:08:52]
And the bus Sunda light rail they don't go anywhere and that they don't go everywhere they focus only on the high density corridors OK so they have very short roads but very high density roads so let me show you case studies the 1st one is this City Hall many many of you new come there are you know straight here OK to people OK so so one of the things that you have to understand is that you know there are 2 cities that are important you know what you know them.

[00:09:20]
Sydney and Melbourne No No So in the US you have New York and Boston right the hate each other OK So Sydney and Melbourne you know take that and multiply by 10 they don't even play the same sports so they really don't like each other so at some point they had to choose where to put the capital and they couldn't agree so they went to the middle of the country in the bush country there is nothing there and they build this city OK so it's a very green city so this is it looks like it was built by an American architect and so you have a lot of different communities there so this of various you know kind of little town with their own shopping centers their own movie theaters and so on own restaurants and they are separated by these a lot of these green space OK So when you think about it is this is like the you know it's a city which is very very widely dispersed OK And one of the things that happen when you do that is public transportation is very difficult to have very long route.

[00:10:15]
The frequency is low because otherwise it costs too much but because of frequency low is low people don't like to use it and therefore nobody is happy it costs a lot of money and people don't use it very much. There are no congestion in Canberra or like 5 minutes a day OK so what I'm going to show you is a new system that we're proposing simulated for that particular city OK and what you see there is blue are going to be the bus and the total they are the more people they are and this is a hub this isn't a hub OK And then the green things are the shuttles moving around and what you can see is that they are really feeding these buses OK so you would stay this is moving like crazy so you see the shuttle moving there and then the buses going to pick these people up so you see a bus where am I buses OK You see this best people are accumulating there the best is coming.

[00:11:03]
Then he's going to pick it up and then move to the next to the next stop OK and the shot to the basically synchronizing OK All right so. This is the cost of the system before and after this this particular model so this is the existing cost in dollars and this is the commuting time this is the cost of a new system which is top of it and the community time which is hard for it so what I want just want to show you here is that for a very widely dispersed region where the population is sparse this systems work very well you can decrease the cost you can improve the convenience significantly.

[00:11:40]
So that's one case study some going to give you another case study which is the city on the knob are completely different so you're going to see a very dense system in a very small geographic area OK So this is a city in Michigan there is a university there as well.

[00:11:56]
As 2 transit system one from the city and one from the university itself and they are very well run both of them so this is the one from the university they have about $50000.00 commuting trips a year to give you a sense Detroit is about $70000.00 OK and he says it's $75.00 capacity you know 55 percent of its capacity which is where both national average once again it's a very dense graphical area which explains that.

[00:12:23]
This is a nice picture which is basically showing people students. At the bus stop and can not get into the bus because the bus is just too crowded and believe me I mean you know that these humans can actually pack these bus like no one else right so so what I'm going to show you is the visualization of this transit system a treatise on time to speed so this is going to go very fast the blue again the total they are the more people they are read all the people waiting and once again the bar so are the number of people waiting OK So this is this is the night so nothing is happening.

[00:13:01]
And so what you going to see here is that they are highly congested corridor here Ali congested go read over there and then you know there is a lot of people moving between these different parts of the compass OK All right I'll come back to that and I'll do.

[00:13:17]
I'll go slower on I'll assume to the system later on so what we did was essentially saying hey these buses we have too many buses we're going to reduce the number of buses they're going to only serve these Corydoras and then we would have Dish and so we showed them the roads and then we have the shutdowns that are going to be going basically feeding once again these hops and going into the extreme heat of the campus OK So this is one of the shuttle that we had it as a boat people this was the shuttle in February last year.

[00:13:46]
So what I'm going to show you is before and after Yeah this is that this is up and into most of us these days OK So this is the existing system 7 pm new A.T.M. and this is real data right real data OK simulated real data and so you see this so we are basically reaping what the system is doing and so you see the bus here very heavily used there very early used here but what you can see here is that on the extreme ease of the campus that basically running empty the 1st time I was on this campus I was looking at these buses 2 people in there I was like wow you know why they're doing this OK Of course if you go here you know you see these buses which just probably about 50 people at this particular point the buses there are crowded Hindustan high density corridors OK.

[00:14:31]
So this is the same system but 7 30 in the evening 7 30 in the evening the bus frequencies twice as long and what you can see though is that a lot of students who obviously don't stop doing things of their 7 pm basically waiting a long time to actually take the buses so know the system diversity some more heavy use but the students are waiting stuff with you waiting a significant time OK you know what I'm going to show you know is the new system using the shuttles and one of the things that you would see is that there would be no red peaks and the only red peaks that you can see are going to be a bus stop but the bus frequency is one or 2 minutes so is very low OK so all the shuttles here are bringing people up picking them up and the waiting time is their own 3 to 4 minutes and therefore you never have somebody waiting very very long OK what you can see is that these kind of busy moving moving people around all the time I'm going to zoom on this because we do automatic right sharing and I wanted to show you that So this is once again zooming on a part of the campus supposed to move and what you can see here is once again the shuttle the total they all the more people they are and you see the shuttles with 2 or 3 people this is probably 4 people this one is probably 5 people many of them ever only 2 or 3 people but this is essentially the system automatically doing right sharing as much as we can or go up to a certain level and there is a one constrains a tween poses that nobody takes more than 25 percent the length of the shortest path.

[00:16:07]
So this is some statistics on the system Remember this is a very high density said this is like fold fold by 4 miles OK and 50000 people traveling every day what you can see is most of the trips are one or 2 legs which means a bus and then a shuttle or a shuttle or a bus or just a bus or just a shuttle OK so very different from Canberra you know sense and you can see that the average time though is about 40 minutes for a trip about $34.00 minutes for waiting time in our fridge so which is quite quite nice compared to the existing system where you can wait 10 to 15 minutes no I want to show you the cost the coast here everything was up to mice to maximize convenience and we wanted to see within the budget at all of the transit agency and so what you can see you know is that we fit inside a budget most of the cost is the driver cost so that they we have automated cars this goes decreases completely we can even improve the service or go further away call for a larger retail.

[00:17:07]
Yeah that's what I wanted to tell you so this Congress will make a very big difference at some point when they come and become available in some fashion so essentially what I'm trying to tell you is that what you are trying to do here to step back a little bit is that we are trying to reinvent public transit so what Ober and lift have done is reinventing taxi services and using technology to make them better what we are trying to do know is reinvent public transit and have system which use technology to completely rationalize transit and make it much more efficient OK once again supply chain of people inside of instead of just you know I think fix roads and fix get it everywhere.

[00:17:49]
This is what would happen if we only use shuttles OK So we have only the shuttles with people we're using them all the time the price is $10.00 times as much OK now you're going to tell me Yeah but this can be automated vacant at some point but this would be what this some of the of the high density Korea are so you would have $200.00 cars there and you would create massive congestion OK so in a sense you need this multi-modal system to actually transport people effectively So let me so this is not new I mean many cities know I've discovered that burn lift of increase congestion in the cities so I think it's 3 percent around 3 percent in in New York City and some cities are basically trying to prevent the number of Oberon lift cars inside the city itself OK So this is not surprising we are just giving you the real numbers if you do a case study like like wait it.

[00:18:40]
Also mean candy bar izing the transit system and therefore we aren't a very dangerous point in time we're transit could actually die and so so I think it's a point in time we really need to use technology to improve transit. So we have deployed this OK so these are some of the mobile applications that we have done and I'll show you a couple of so the only reason I want to show you this I mean this is like we're going live on steroids but the only thing that I only think the reason I'm showing you this is that we have this concept of stops I told you that we have to pick up people very close to where they live but we don't want to pick them up their home because the older we loose time when we do that they would for instance you know forget their keys or they want to kiss a wife so their husband and they take time so we don't want that we want them to be ready and we also want to come to a point where we can actually do right sharing so we have this concept of you have to stop where people can be picked up together and we just you know can pick up many people at the same time they are ready when we come.

[00:19:40]
When we have already spotting system is like a cloud computing platform not not very interesting this is a visualization of. Of the money tearing up the occasion so we were watching this system you know every every day that would drive my wife completely crazy but actually after a while she was watching with me because this is very interesting you see all kinds of interesting things so this is Kelsey This is one of a driver so you going to see she's going to do something very weird but when we click on the drivers we know what they have to do when you click on the people who are waiting we know where to go and so you see cats you know Kelsey there as to so she's supposed to pick this person up and to bring that person to her residence there but somehow she never knows that she has to turn right there she always go this long way and initially that was driving me completely crazy because I never knew that she was actually doing the right thing but we're going to see she's going to do the wrong road and obviously Google doesn't know that there is a threats to this parking lot this way but she's going to do that very interesting so we discover many things by watching the system and we made a lot of improvement to the algorithm just watching the pathological behavior that you could see so you're going to see Kelsey is going to go up yes.

[00:20:51]
OK So one of the things that we notice I told you that the 1st one so I told you a couple of things at the beginning that the 1st my problem was an issue I also told you that Groceries were an issue this is where people go inside the system so most of the transit system stopped there so we expanded the region of coverage and you see these points they're very heavy used very very very popular destination the residence of stuff and faculty and and students obviously and so a lot of these students had no way to actually get to the campus.

[00:21:25]
Except by walking and then taking the bus off the words all taking over and left a lot of them would do that and so on this is very interesting this is not a dorm this is actually. A plaza where they risk Roeger and a number of be a places and you can see that this is highly popular as well so a lot of the students who are actually using the new transit system for actually shopping and for regrowing to get a beer or 2 or 3 I mean you don't drive any more so.

[00:21:54]
This is the same thing hit me up so this is a distribution of the right most of them are going back home or going back to school IT percent our grocery one percent is health care they would go to a have different city using the transit system. This is a waiting time so that's what 2 different this with different number of shuttles but like you can see here is that most of the time here where you know waiting time was less than 3 minutes there's some 5 minutes in this particular case Now some people are waiting about 10 minutes that's typically when the system starts or when the system finishes because everybody rush to actually use the system especially at midnight when the system stops everybody wants to go home they play so many requests that sometimes the system has to go up to midnight and 20 minutes or something like this.

[00:22:39]
Beyond a scene I'm not going to talk about this too much but this is what we're doing so we start from the road data what people are people what people are what people are expected to do and we use a lot of data science and machine learning to find origin destination over time and so no we know what people are doing and therefore we designed a new system and then we have essentially the roads and the number of shuttles and so on and then we do the real time operation which is really time Dayana right problem OK so.

[00:23:06]
Yet and once again there we use an optimisation machine learning to up to rebalance the fleets and do things like this. This is a little bit whole we do the planning so we do a network design and then we do trip splitting we set X. is the different zones for the foot of shuttles we if we fleet size them properly and then we do the rostering of the drivers we control the Reivers here but they have shifts so we have to obey the regulations OK so that's all the optimization algorithm that we have to solve so it's kind of cute.

[00:23:35]
This is one of them I just want to flush some just you know just just for fun. Splitting is interesting so this is at any point and we do that in real time as well you have somebody who wants to go from one place to another and we try to find the most attractive trip at this particular point in time so we try to find here which is the best hub you know depending upon where the bus the song when they are going to come and so on and then they go to another up and then we have a shutdown for instance so we optimize in real time we already pre-computer a lot of things but we optimize in real time to find what is the best trip at any point in time and so this is a nice resource constraints for the spa so it's very nice of the musician problem as well.

[00:24:15]
So let me take a step back here because I took a lot and I want to give you some perspective here from and historical standpoint so so I told you that you know we are trying to do for public transit and lift it for for taxi services and I told you that if you only use and live to have this big big issue.

[00:24:34]
And so so I want to go back a little bit in history and so this is a Spanish writer. Who said that those who are ignoring the possible to repeat it are doomed to repeat it and so you may want to know who has invented public transportation actory anybody knows nobody knows right can you guess when this was invented.

[00:25:03]
Not really no to really notice this purest form OK but you're very optimistic MARTIN You're very optimistic this is a this is this is somebody who is a nice name. And so this is 1723 so great mathematician OK and I never knew this OK I had to do some research to actually find this out and so you invented this carry it in 162 and it basically run for 15 years and then he stopped and I tell you I will tell you why it stopped in a moment when he was very popular and then had to restart it about 200 yourself the words in Paris again and so this is this is the you know really the nice the nicest places in Paris and they had this these busy roads to actually be servicing people you know this is Hollywood public transit right so you had 2 horses and then to carry it it was very you know very interesting and that would run every 15 minutes and so essentially what they said it's very interesting when you look at that his story of this thing what they are telling you is that public transportation had a big effect on the life because people could goo and go beyond their neighborhoods so it was again accessibility so everything about public transportation is about accessibility giving opportunities to people and so I did that stop it stop because the ferry increased the price increase and then the 2nd thing that happened is that they put regulation and only the rich people could use it and then the rich people preferred taxi and public transit.

[00:26:31]
OK So interesting lesson from history right now the tax so one of the things that you saw over time as well is that most of these public transportation were very very quick to adopt new technology like you know. Steam railway and things like this. Taxi started to a wrong the same time which is very interesting or both at the same time in England 1st and then in power is roughly the same time as public transportation so very interesting using exactly the same technology OK And then of use lease all taxis in the beginning in New York City.

[00:27:05]
Or on the beginning of the 19th century the end of the 19th century and then you know we got this and then we got this OK So essentially these 2 these 2 system evolved very much in the same fashion except in the last 10 years in the last 10 years who got a lift and make a tremendous difference but public transit does not move very fast so there is this big gap between the 2 and so we can think about that right so this is interesting.

[00:27:30]
OK So let me talk about a completely different different model No it's a completely different mobility service and I'm going to give you the most boring 3 D. video that you ever seen in your life OK this is so we're going to see so this is 15 parking lot in one city OK the most used parking lot in that city and the toll of the bars the more people they are in this parking OK so the number of cars this is the number of cars and you can see how did the morning you know this is increasing up to Iran 99 am and then nothing happens so it's a treaty video that doesn't move.

[00:28:02]
Because these cars are staying there forever OK for these 10 hours you know a parking spot parking like this cost about $40000.00 so a parking lot like this is about $40000000.00 OK And so that city is actually a lot of issues with this parking lot there is a significant pressure on parking know people you know taking these gaffes and going back home so essentially you know these kids these park these car said been there and don't do anything and the and they take place in this nice parking lot which are very pricey pieces so very distinct OK So so the person running this parking lot came to us and say can you solve this I mean can we actually avoid having all these cars in this parking lot and so what we said is OK So let's look at people who come to this parking lot and view them as communities OK and try to see if they can come and write Chair OK And so we want to minimize the number of gaffes in the road OK And we know when they come and when they leave OK so they so we know exactly when they come to the parking lot when they leave and therefore since we know also where they live we know when they depart from home and when they go back to home OK when they go back home and so what we said is and so we are assuming that we can vary this by let's say 5 to 10 minutes OK so so and we wanted to make this very big trick sharing experiment OK You know one of the reasons how sharing and calculating doesn't work very well is that it's very difficult to match you know colors and people very very difficult.

[00:29:30]
And so there is so there is a very interesting thing here can we use actually good data science and up to musician to actively do that and that's what we try to do. Before doing that we said he what is the property that the platform like this has to satisfy and I'm going to show you a bunch of them if you look at the social science and various transportation Journal they will tell you that you need the writers to be especially close such that they know each other obviously they have to say the same schedule and if they come in the morning you want them to come back in the evening obviously the F. to have a guarantee right because they will never use the system OK so they select 33 properties that you cannot violate and then the next one the next 3 ones kind of desirable property if you want to century low coordination cost you'd like to travel with the same people in the morning and evening or every day of the week you also want.

[00:30:23]
You know it's the same people the you know you don't have new people coming in and then you want also that if you drive if you want to drive every day or be a passenger every day you don't want to change the role of the time and so what we try to do is build a platform to satisfy as many of them OK so this is the this is this emission of people all over the place to this is about you know 50 this is about 50 miles.

[00:30:48]
This is Canada this is the only place in the U.S. we kind of ice off of the of the U.S.. This is when they come when these people come so this is the entrance in the parking lot and this is leaving the parking lot and you can see that people do very very predictable to come at the same time they leave at the same time except this particular cluster there and you guess what that is lunch no no no this is some people call me go to Iran 67 and leaving at 4 am No no student student students do that but not that many of them.

[00:31:26]
Night shift night club no. Well you guys have interesting priorities know so this is the night shift of the hospital OK So nurses doctors and patient while mostly nurses and doctors so this is a this is a slight I used to sit up people are boring people told me don't do set up so people are very predictable every day of the week they come around the same time they leave or on them all the same time there is a little bit more uncertainty when you leave in the afternoon the spread is a bit bigger OK but this is amazing right me but I strongly predictable and so we try to business platform and we've been various algorithm implementing as many of these principles as possible OK and I'm not going to tell you how we did this I'm going to jump to the results because this is the most important part so they saw the results of the various auguries this is 100 cal there is no sharing this is when people are commuting every day with the same people in the morning with of same people all the time the same people you save 2 percent which is nothing and the reason once again this is this is explains why the scallop pooling program some of that successful So what you need and in this particular case of people outside the city you can centrally remove her from the car if you are a low flexibility if you match people in real time every day in the morning in the afternoon and every day different so you need to make this successful this is inside the city you have more opportunity for sharing so you can essentially remove a tutor of the car OK Once again what you need there is the flexibility you need to up the mice or the time behind the scene what do we do we have a 1st clustering algorithm this is the clustering algorithm that you see there we this is through to I have a special look at Lety so we cluster people in different in different regions where they live close this is inside the city itself.

[00:33:19]
And then we basically do a big matching algorithm for rider and and and and and car so the key point here is that you have to make sure that you know they have a right back which is this constrains here so if a car is coming in to test to go back in the evening you have a flow balance for these cars.

[00:33:39]
So you minimize the number of drugs you serve every customers involved same driver same Bawn an oddball sort of this is the main concerns that you have to satisfy What's missing is so you compute these roads I'm not going to tell you what what what ho we do that does the most interesting part because that's that's that's where the science is located but roughly speaking it's like another resource constrained problem with a lot of different constraints.

[00:34:05]
Yeah this is them up OK No no. I told you that essentially what we are doing is kind of a supply chain of people and people are little bit different from packets OK And and so this is this is a I'm going to show you some statistics here for telling you you know some interesting findings that we had not only you have too much people in real time but you can see here is that if you increase the capacity of the vehicles instead of living you know a council for people if you have a car with it people so what is what is happening is essentially nothing OK so most of the van you know you have a lot of fun putting says you know programs there but really you can reuse this van So what is the big difference is between 2 and 3 and 4 but after that you essentially no benefits there you can see that so increasing the capacity of the vehicle here is not going to do anything is very difficult to measure the people even in one car.

[00:35:01]
This is also the impact the average or the average. Ride duration and what you can see is that when you move from one to 4 you only increase the travel time by about 14 percent that's very very low so people actually dicom you would increase by 14 percent which is almost nothing right and so.

[00:35:21]
So that's a very good indication we have a concern which is we don't want to increase by more than 25 percent but in fact in practice 14 percent is the average and this is the this is the impact of the cluster if we have bigger and bigger cluster you can see that we save you know we can reduce the number of car significantly but you have the musician problem become very very complex at that point but you know the one the one way you can benefit is actually try to point to increase the special look ality of the people that you're matching.

[00:35:50]
Yeah so the same thing this is the performance of various algorithm most of the time you consult I mean the real time so less than 10 seconds if you use this you know yellow This is yellow or green I don't know I'm colorblind so whatever that colorist that's the best algorithm that we have it takes about 10 seconds most of the time there are some instances some clustered other little bit more difficult so so one of the things I want to do know is again step back so why is this so difficult and I'm going to show you a neighborhood so this is a neighbor route in the city that I've shown you this neighborhood is about one miles from the parking lot right so you have these people this is about 400 people 500 people they are commuting for one mile leaving the house for 10 hours in this parking lot this is all of them this is where they live OK And so what I'm going to show you are 2 graphs OK one whole condition all much they can share in the morning and almost they can share in the evening OK so the ground that you see there it's highly dense right so there is a natural thing to people if they can share in the morning right so it seems that a lot of people can share right nice so this is indeed evening very dense a little bit less I told you that this is a bit less predictable.

[00:37:00]
But what is interesting you have to have the right back right so what you want in practice is an intersection between the 2 and that's what you get much sparser that's where the problem is much more difficult so you have too much disk in the morning in the evening and when you get that you don't have that much sharing and that's what you need to optimize OK this is a simulation so this is a bit boring because this is 6 am But let me move to this is not this is not a plant right so.

[00:37:29]
It's not as crowded but you can see all these blue coats coming inside the center of the city. All these people sigh commuting alone in their cars OK now I'm going to show you when you right share OK So this is. This is this one OK so we're going to see you know and the interesting point is that you have to look at these bakers this blue and then they become pink and then they become red blue is one person pink is 2 or 3 and then red is is is is 4 of 4 for people in the cup and what you can see you know is that most of the car rental pink and if I move to AM you're going to see a lot less car and a lot more red you know red cars there and so you reduce tremendously the number of couse in the.

[00:38:13]
In the in the in the city you know one of the things that you have there is the best possible thing OK So this is. This is assuming that everyone to want to go right here and one of the things that we don't know is how much people are willing to walk to the right sharing practice even if the increase only the travel time by 15 percent so we're currently doing various kinds of survey analyzing them to actually find out once again this is a difference between pockets and actually people you know that they can choose to right here on the OK so let me talk a little bit about a real time generalized Dayana right because this is of the core of everything that we do so this is New York City this is all the trip the taxi trips for New York City in a particular day so this is origin and destination there are 40000 a day if I remember correctly.

[00:39:01]
And so what we do is we try to see if we can actually optimize the system using the using right sharing again so it's a taxi services but automatic right sharing but we don't want people to spend more than 25 percent of their time compared to the shoulder spot and so what we do is we budget the requests we optimize we budget the requests we optimize every time and we try to maximize the minimize the waiting time.

[00:39:25]
So can we do this and what would be the quality of this OK so this is the optimization problem at the high level once again I have to show you all to generate the roads which is the core of the problem but this is essentially minimizing the minimizing the wait time and then penalize ing people who are not being picked up OK So over during every period if somebody is not picked up the priority of the person increases that at some point of the person will be will be will be served so it's like over in life but with automatic right sharing and guarantee right so nobody is going to turn you down anything you know we are not going to turn people down and so.

[00:40:03]
Yeah and we have this constraint that you don't travel more than 25 percent of your shoulder spot and so this is the density of the requests doing one particular day and so this is an interesting thing that we can do so Currently there are about 1000 taxis in New York City so we can optimize everything we have 3000 vehicles and a capacity 5 and then the average which is one minutes and 30 seconds OK No I don't think over and it can do that so they are like 10 minutes at least the maximum width was $28.00 minutes this is one of these peaks that you have this is during rush hour we optimise by using rush hour in the morning and evening so this is what we can do when you do up the musicians so this is pretty cool I think OK so once again this is a the core of everything we're doing these basically these generalized dire a right problem OK beautiful optimization problem.

[00:40:49]
So conclusion so what I want to tell you is that we are building this this Caleb and mobility system which I really view as like supply chain of people. And. If we are successful we can have a fundamental site societal impact because mostly we are talking about accessibility given access to people to various services and you know including jobs and health care and so on behind the scene this is an area where we amazingly interesting problems in data science and in optimisation and machine learning and this is the beginning so what we are doing though is trying to fix up and that's what I would like to do is going to take us a many many many years because well because.

[00:41:30]
It's most congested city in the world which is pretty bad right and a commuting patterns are very interesting so they are 85 percent of people commuting alone or go and the public their share of public transit is only 3 percent if you have a city like bust on I think it's about it if you know 15 to 20 percent so he's very very smart so the transit system here is not very much use and the difference is once again a big factor between driving your car and doing transit so for me this is the time that I would that I think by car if I think transit is about 80 minutes at the minute so it's much more than active at this this is an average So this is a very big challenge and this is what we are looking into so we are collecting all the information about you know what people are doing in the city and then trying to see how we could fix this system and what it would take to fix it so that's that's what I wanted to tell you and I would be happy to take any questions if you ask them thank you very much thank no question.

[00:42:56]
We had the data because we we most of the parking lot here including here and so we're getting that data for Judge attack and for the city itself so people if you go to a parking lot you typically swipe your car Ideally you should fight broke out when you get out so that's the only thing that we need and then the address of these people and that's the only thing that we had there so for the community based sharing it was very easy because in a sense we had the rest of people and we know when they would come commuting in on not know what we didn't have is that do they have to pick a big hits on that and that's a big does a social part of this so so some people will not care share because they have to pick up their kids and they want a car and that's part of the commuting so one of the things that we have to find all these among all these people what is the percentage of these people who are willing actually to cash it so that's a big thing that we have to learn and also incentivize people in the right fashion yes.

[00:43:57]
No you have only the driver the only thing that you have as a requirement is that the most flexible system that we have is that the driver has to be the same in the evening and in the in the morning in the evening that's all really quite they're going to be different on different days otherwise you lose about 20 percent so I think the flexibility of the driver is also very important.

[00:44:18]
Yes yes I think this is what we're discussing we've various in various in various in many of the project we're actually discussing the right incentive so a parking spot typically cost a lot of money OK so this were actually a very expensive parking spot there and so we were basically trying to incentivise into is one was to say hey you don't pay for parking anymore or you will reduce the price of centrally the other $1.00 is that we give you very nice spots like you know very close to where you park it where you live where you work and so on that's the kind of incentives so that more and more I think incentivizing people he's a very interesting and very very very nice topic and the pricing of these things is also very interesting so I think this is this a beautiful area that we are starting working on as well yes.

[00:45:12]
Yeah. So. Set up again so you're worrying that people are cheating on them yeah also if they were one to do that if they were good at all but if they would do that I would be very happy we find another incentive I mean I mean remember parking a parking lot like This Is $40000000.00 so I have a lot of money to play with right so you can think about this that way so I think I think if that happens does the ideal situation but you know I can tell you that there are people who will never share so we presented this to the higher administration and the Dean would say we were never sure any of anyone right so it's very interesting so people would never share they won the space for themselves or they just don't want to have somebody that they don't know even here they would know people right so that but they just want the complete flexibility so I don't think we'd get to there but if we remove the number of by 20 percent for me it's major right so it's like 30 percent is a big big reduction if we can do that I would be I would be very happy 30 percent reduction in greenhouse gas emissions would be fantastic on a question yes.

[00:46:36]
Yes So in many of these transit system you don't get everything so you basically typically get when people are bored and not always because some people pick up what you have you have a record of the transaction you have some you know most of the time you hold the boardings but you don't this is City of you know I think even in even in Mata if you take the train we know boarding into lightings if you take the bus we only know boarding so we have to do inference of who do we do these people don't like so that's one of the problems that we have so we build these probabilities distribution which use a number of assumptions because typically people would lie very much the same way and so if you were light in the morning you probably born in the evening from the same position so we use assumptions like that to try to predict as best as possible these are reaching destination.

[00:47:20]
That's one of the things the other thing that we try to do is when we design a new system like this people people with change remote if the system is very attractive they will move from owning you know owning big powerful driving their cars to actually using the transit system how many are doing that so we try to learn that it's done this is fairly easy to do but that's what frankly.

[00:47:41]
On a question yes yes so that's what we are studying right now so this is a this is a very good point so one of the things that we are trying to do is all much flexibility do we have to actively shift the schedule is somebody somebody has to leave early or you somebody say I have a meeting my meeting is going to be over so we're studying that exactly not because this is we want to find out or so you can match people in a different fashion to actually maximize flexibility that's what we're doing but you know the optimization problem is a little bit more complicated to say the least you know just to it's more complicated yes.

[00:48:23]
Yes Friday is a different day so Friday is a little bit different day so you have mostly the same production but just as impressive because there are fewer people driving so when people take vacation or one day off that typically take Friday and so there are fewer people on Friday they have also stated which is much more difficult to predict a little bit so Fridays special So Monday to Thursday is the best days most of the week so similar you have a little bit difference in the 1st and the last week but mostly a decimeter but Fridays are Fridays a special day yeah a little bit good you watched everything yes.

[00:49:03]
Because we. All do you are not in the same vehicle in the most flexible you are not in the same because you're in a different vehicle but you have one vehicle so so what we want is the we want a vehicle stop from one location and go back to the 2 occasion.

[00:49:18]
That's the only recourse right so what are the same person Dr or not that's not even a contract I mean there are one percent difference it doesn't matter so you could do that we both made it callous but what you want is a neighborhood to say well to do you have a limited car you can relocate some of these cows so that would be very interesting to see what you can do but currently you know the carrots have to go back to the same community so that if you're limited cows you're going to rebalance with the magically Indeed it would be actively I mean.

[00:49:50]
Absolutely and that's what people do most of the time they are not in the same car but the car the car is to go back to where it started on the neighborhood where it started the clustering started so you so you have a lot of like you have almost every flexibility except that you know the cut the number of counts in every one of discuss Teresa same at the at the beginning of the day and the end of the day now.

[00:50:19]
So I think the shuttle the shuttle is very interesting so the shuttle ascent of average about 2 to 3 people in average they'll peaks where they have about it people obviously Dave people seem to peace but the average His none very high is 2 to 3 but the shuttle some making a big difference I think that you have it in the on the man transit system I think it's eighty's a good number is a good number is more difficult to do it to get the same results with 44 cars because here we I mean the it's a multi motor system so if you if you try to have only car then it's so light like we did for the taxi service this is very different but this is also very different I mean one of the things which is different is that the transit system is going to do commute while and left on no the really doing commute they are doing other things and so very few people use aware and live to do commuting and so this a very different situations and big and therefore the data is much more sparse in terms of the number of people that you can actually.

[00:51:19]
Put inside the same cow for a system which is not about commuting That's why there is a difference between the 2 systems but enough bridge it's depressing a whole empty these troubles but you need them for the Speaks and the peaks are where you know so the system is up to my sense that you are very good for the peaks otherwise you would have people that are not serve well not so for about 60 minutes and you don't want to have that yeah.

[00:51:49]
Yeah. Yeah. I'm sure I was true. Absolutely. Yeah. I completely agree that we could do that I completely agree so one of the things that we found very interesting is that a lot of the a lot of the people who are actually using the new transit system to go eating in the evening to go shopping and so on and so you could also use these particular shuttle to actually bring food to them instead of them going there so I think there are tremendous opportunities there so what was interesting to me is that everything that we speculated that people are going to use that for grocery they're going to use that for health care they are going to use that for 1st and last month actually were true so this is really an issue and you can really address it I think that's what I find very interesting and obviously I completely agree with you I mean I told you before I mean it's like a supply chain that what we are doing is basically ideas from supply chain but but it works for people and this is very different from what people are used to and for us to keep the keys is the phone and the technology the communication technology on a question Great thank you guys thank you very much.