This WebEx call is being recorded and may be posted on the website or used internally. If you do not wish to have your voice recorded. Please do not speak during the call. If you do not wish to have your image recorded. Please turn off your camera or participate by phone. If you speak during the call or use the video connection, you were presumed to consent to the recording and use of your voice or image.
Good afternoon everyone. Thank you for joining us. I am happy to welcome you all to our third in the series of webinars to discuss the results and insights from the first phase of DOE’s Smart Mobility Lab Consortium. My name is David Anderson. I am the program manager for Energy Efficient Mobility Systems, or EEMS program of the Vehicle Technologies Office part of the Department of Energy's Office of Energy Efficiency and Renewable Energy. Our webinar’s take place every two weeks. We started on September 24th, and we plan to continue them until December 17 each one focuses on a different aspect of what we accomplished and what we learned through the smart Mobility consortiums research over the last three to four years. In our first webinar, we discussed how the smart Mobility modeling workflow was built and in our second webinar. We learned about the development of the mobility energy productivity metrics now today, we turn our attention towards results from these tools.
Our Focus today is on the application used in the results of the smart Mobility modeling workflow. We have two different implementations of this software tool chain built around to different agent-based transportation system models and we will hear about how each one was applied and exercise in just a few minutes. But before I turn it over to our speakers from Argonne National Lab and from Lawrence Berkeley National Lab, I will again give you a brief background on the Consortium. EEMS is one of several research programs within DOE’s Vehicle Technologies Office or VTO. VTO conducts research and development of advanced vehicle and transportation technologies that improve the efficiency of our vehicle fleet and of our overall Mobility system VTO supports component level research on Advanced combustion engines and fuels, advanced materials energy storage and electric Drive systems, and this research is done in the context of the overall vehicle improving the efficiency of the vehicle Itself.
Now the EEMS brings connected and automated Technologies into the mix and we researched these Solutions not only at the vehicle and powertrain level, but we develop Technologies to improve system level of efficiency at the intersection or network scale improving overall traffic flow and ultimately working across all passenger and Freight modes to develop solutions that improve Mobility at the entire Urban or metropolitan area. To do this the comprehensively research in model and develop solutions that reach across mode and domains and take the multidisciplinary approach. We must consider Transportation holistically as a system that is a highly interactive highly dependent very complex system of systems and exchanges in one area propagate to and affect other areas. So, what we must understand each of these systems individually is just as important to understand how they fit together how they interact and how they affect one another.
To integrate these systems these modes these domains, we built a multidisciplinary Consortium we convene five of our leading National Labs doing Transportation research. We pull together Transportation scientists and Engineers expertise and vehicle modeling and simulation testing traveler Behavior charging infrastructure, freight operations, Urban science and land use we pulled them together to form a multi-year multi laboratory collaborative that we call Smart Mobility.
This is a cornerstone research effort for the EEMS program and is focused on one understanding the system level impacts that emerging Transportation Technologies and services will have on mobility and to identifying solutions that improve Mobility energy productivity.
Through EEMS and through smart Mobility. We ask ourselves questions. Like “what does future Mobility look like?” “will it be fully automated?” a lot of researchers and technology companies and all the manufacturers are working towards developing a fully automated Vehicles. Some forecasts like this one predict that the share of AVs will grow tremendously in the coming decades with a third of vehicle miles traveled being served by AV Services behalf of the until you could be served by personally own of needed.
As newly automated driving features are introduced, we must How they operate not only how they impact the vehicles equipped with them. But also, how they affect surrounding traffic and how they affect travel to me. We asked ourselves as new Mobility Services reach maturity. What does this mean for how people travel? What does it mean for the number of cars on the road for curb management? We have seen a huge increase in transportation Network companies in the last decade and largely at the expense of traditional taxis. So, what happens if the growth in Mobility as a service continues
We asked ourselves in addition to new Services. What about new modes? For example can micro Mobility actually have an impact Our Lives affect other modes such as Transit, you know, it's easy to disregard bikes and scooters, but we see the trips on these modes more than doubled from 2017-2018 and they grow Again by over 60% from 2018-2019 to a hundred thirty six million trips.
And finally, e-commerce, this is a big one here. We have seen a steady shift from brick-and-mortar retail to online shopping and delivery over the past decade, even before the impacts of COVID-19. And so, we ask ourselves to what extent will this trend continue especially given the cut discontinuity that we've seen 2020. What impact will it have on our transportation system in our new and future modes like package drones efficient. Are they affordable? Are they even viable? To tackle these complex questions the smart Mobility Consortium created end to end multi Fidelity modeling workflow built around agent based transportation system models that capture the complex interactions among decision making technology implementation different Mobility service models and those land use and EV charging infrastructure.
It allows researchers to evaluate the energy and mobility and the affordability outcomes of potential future Transportation scenarios, and we can evaluate these outcomes in terms of vehicle miles traveled travel time energy cost, greenhouse gas emissions and other metrics. We think this is a critical research capability, but the real value is in working with stakeholders. That is the transportation planners the technology providers local DOTs policymakers and other end users. Our plan is to ultimately deploy this Capability Beyond a research tool, so we invite you to let us know if you're interested in partnering. You can see our contact information here on the screen. So now I will turn it over to Josh Auld from Argonne National Lab to describe how they implemented the workflow based on the Polaris model in Chicago. And he will discuss the results of they did they generated there. Now after Josh. Zach Needle from the Lawrence Berkeley National Lab will describe the implementation of the workflow built around the beam Transportation model and applied to San Francisco. And once Josh is done describing the results will have a few minutes for questions and answers for encourage you to all type your questions in the Q&A box that you see on your screen at any time and we'll go through as many of the questions as we can as time allows.
So now I will turn it over to Josh Auld from Argonne National Lab. All right. Thank you, David. My name is Joshua Auld, I am the technical manager of Transportation Systems and mobility within the vehicle and Mobility systems group at Argonne National Laboratory. And as David mentioned, I will be discussing today the application of our Polaris Centric version of this normal Mobility workflow to analyzing scenarios for the Chicago metropolitan region.
As David mentioned, this permeability workflow is designed to be a comprehensive approach to addressing complex questions around Transportation Mobility energy use future Technologies and so on. The version we will be discussing in this presentation is centered around the Polaris agent-based transportation system model. This is a mesoscopic simulation of an entire Urban and Regional areas that's focused on modeling traveler Behavior system controls, goods movement and how people travel now this is the Centerpiece of this Polaris workflow but it's also connected to a number of other modeling components from long-range infrastructure type issues around eat models of EV charging land use changes driven by Transportation changes and vehicle markets to more microscopic Des aggregate type modeling around microscopic traffic flow models and multi-vehicle control models all of these feed in and connect through the workflow to generate trip profiles travel realizations and then vehicle energy consumption. Which is used to generate the mobility energy productivity metrics. The discussion of the previous in the webinar series, which is a way to compare changes in transportation mobility and energy use in a single metric evaluation. This workflow has been implemented both by Argonne National Laboratory and Partnerships with several other National Laboratories, including Oak Ridge Berkeley NREL and Lawrence Livermore as well as numerous University Partners in terms of getting all this work into place.
So why do we feel that the Polaris workflow is unique? There are a few a few key reasons one is the modeling features that we have implemented as both within Polaris and in the connected tools. So, Polaris is a full-featured activity-based model that also includes Freight shipments and local deliveries. So this means is an integrated fully integrated agent-based model of travel demand number of assignments and traffic flow all operating any continuous environment that also is connected through High Fidelity vehicle energy consumption as was mentioned in the workflow as well as EV charging and pretty integration additionally our connection to the urban sin model allows us to forecast changes in land use driven by changes in transportation performance this the focus on activity based modeling as well allows us to account for travel or behavior impacts driven by changes in trouble time valuation of trouble time due to technology.
These are no ability systems across many of the choices mentions that derive and drive the travel Behavior individuals. Now, this is enabled by a focus that we have taken in developing the Polaris in the larger workflow on computational performance. So, one key factor is that the model is fully agent-based or representing all of these different decisions for all of the different travel agents. It is also integrated with external optimization tools and solvers things like cplex and Keroppi to solve large scale multi-agent. Problems things like optimizing trade flow and TNC assignment and so on all of this is enabled by the development of the model itself in a high-performance C++ code base that is cross-platform and implementation independence of this can run on both desktop environments and cluster Computing the HPC environments this focus on high performance Computing high-performance C++ coating allows us to run large scale models with up to a hundred percent of agents for example in the Chicago.
Our model has 10 million Unique travel agents in about 6 hours of runtime. So, the workflow of discussing now is centered around Polaris and the pleura CBM has needed to be enhanced to account for a lot of these future Mobility Technologies. We are looking at in the various scenarios that we run. So the base Polaris model was really focused on long term midterm and within day choice and travel behavior of Agents the long-term things like where people choose to live and work in which vehicles they choose to own how those long-term choices constrain midterm or routinized activity choices, like working patterns and is going patterns and
Meeting and so on and then how both of those influence the within two choices. So, this is the actual realize travel behavior of how people generate their activities that they're engaging in during a day of those activities are planned in terms of where they're going to take place with both. They're going to use to get there and what time and so on how those are scheduled and consistent activity schedule that can be realized by the person in the transportation system and then how they are routed and actually moved to the transportation system itself. So how which routes that people choose to take and then how those routes are real? Caused traffic flow model and this is a continuous feedback and integration Loop between this planning scheduling routing and realization of the trips. So, the actions of each agent say, they're moving through the system continuously impact the planning scheduling and routing of other agents of this continuous integration between all three of these elements. Now, this has been updated in terms of long-term choices to look at a lot of these future Mobility Technologies. We have had it had in a CV and other EV technology Choice model and a household be able to suppose a model to account for those
It had to look at more of the activity planning Behavior. So how does the changes in value of travel time driven by different Mobility Technologies influence the choices that people make about planning their activities as well as how does it influence the scheduling of individuals to take advantage of some of these vehicle Technologies, especially things like household vehicle sharing for automated vehicles and transferring those Vehicles back and forth between household members. We have also looked at integrating a lot of the Freight Logistics and commerce environment. So, looking at the impact of this. Events on Esther events and such as versa so accounting for growth and changes in e-commerce engagements and how that drives changes in Freight Logistics and then interacting that within the individual with the day choices people make is also going to focus and finally a lot of work has been done on changing the way the system is stimulated whether it's Transit simulation how vehicle transit buses and other vehicles move in the traffic stream or it's more on optimization based of looking at it.
TSA the operations of these are managed rides are assigned and so on as well as updates the core models around study traffic assignments and making that multimodal and time dependent as and capturing the impacts of different vehicle Technologies automation Technologies and changes in traffic flow modeling. So, all of those enhancements have been made to the Polaris model. We also want to validate that this model ensures valid this Baseline model of our model region to ensure that the model has represented in a second. Can capture the Baseline? Chicago so we've done extensive validation and continuous validation business model has been developed a couple of examples comparing outcomes from the activity-based portion of the model to a local observations from seen applicable planning agency are shown here on the left. We can see the trip links distributions and on the right the boats are distributions over area type in the region and both Magic by closely to the observations from the household travel surveys. If we look at more Network level up two delegations we Can see here a comparison of the physical Network loading curves of the vehicles The Continuous vehicles and network curve shown on the left of this is the vehicles in motion at any given time throughout the day and that's compared against a version of that curve generated out of the cement travel survey. And on the right, we have the simulated traffic counts on the highway Network for individual links with the terrorists in the actual Chicago Highway have a network system so that
Detectors about 800 some out of these highly inductive Loop detectors that we are comparing against of the diurnal curve of the volume comes over all those detectors for the same links in the model of shown on the right and again for both of these we met quite closely to the ground truth observations outside of Baseline validation. We also want to ensure that the model is sensitive and suitable for forecasting, so it does not do just to represent what the Baseline conditions are. We also engaged in a process of doing that Costello. Asia and one of the key changes in the Chicago region that was of interest is to see if we can capture the impact of the deployment of Uber and Lyft as a mode in the Chicago region starting late 2013 or early 2014. So, we ran three scenarios both our 2018 Baseline scenario and then whining that back to show that we can capture this observation so we can see the change in observe Transit ridership and TNC ridership over time and the
Simulated ridership counts are shown in the chart here that never quite closely as the work for confident that we can capture these changes in due to write your modes. So, we take this validated Baseline model for our Chicago region and want to use it to consider what happens across the number of future Mobility technology scenario. So, we want to look at three main scenarios here one is a more midterm or medium-term scenario where we have is good extension. Ian of current day Trends we have more sharing or partial automation so high how to model vehicle automation engine is on the highway and more people using right hail as mode and I'm going to look at two more long-term scenarios, which are more focused on high levels of automation. So not just the highway partial automation, but then these two cases we have fully automated vehicles and one’s scenario were calling Sav for shorthand. We have high sharing. So, these variables are only deployed within share vehicle fleets and
Other hand it was the private TV scenario. These vehicles are mostly although some of these Sav fleets exist mostly owned by private households. And these are all compared against baselines and I should mention for each scenario. There are two vehicle technology cases. So, there's a business as usual which is basically a continuation of existing Trends and vehicle technology improvements and then attack success case where vehicles are more efficient. The video program is successful in getting up more efficient vehicle Technologies and pushing out some more of the automation Technologies as well. So, these scenarios are then all compared against appropriate Baseline. So medium-term and long-term changes in demand without the changes driven by the scenario assumptions.
And high level we can see some results coming out of the comparisons. We made across these scenarios. And one of the key findings is a shared fleet in see ABS enable a high system efficiency when we compare against personally on see ABS. So this is looking at the personally own scenario when we when we find them we run the scenario for our private AV case we see about a 22 percent increase in energy use and a 25% increase in all total miles traveled which leads to an 18 18 percent.
To being so significant increase in congestion in the Chicago Network. This does have some upside in terms of seven percent increase in productive miles traveled. So that's all person miles and all Freight miles traveled and that's driven by a lot of changes in the valuation of travel time in these automated vehicles, but it does come with some significant downsides in terms of EHT energy use on the other hand. If we look at more of these shared fleets as scenarios, we can see that overall energy is reduced by about 23% Scent and VMT is reduced by 18 percent leading to increase in seventeen percent. All while providing about the same level of mobility in terms of the measurement of productive miles traveled as the Baseline scenario. Now the difference between these private TV and shared a be scenarios is what a lot driven by operational difference between how these vehicles function in the system. So, we see in hi cheering. Hi ASAP. Hi automation scenario. We have an efficient shared autonomous vehicle Fleet. So, it's a large Fleet that can efficiently reposition between rivals. Questions
Quest We Percent of total miles and we can see in the figure is here the temporal distribution of those loaded and unloaded miles. So about 4% of total the MTR unloaded. However, when we go to the private AV scenario, we have about 15% of total miles traveled are unloaded. So there's a lot more travel and then a lot more of that travel is unloaded due to the inefficiency of repositioning between household members and that has the effect of reducing the temperature vehicle occupancy for a private vehicle from about 4 and 1/2 in the region to 1.2. So, this is sharing benefits of the sharing the benefits. We see in the shared vehicle Fleet scenario is really enabled by efficient. Right he'll operation so this wouldn't be the case if there was a lot of this deadhead PMT that we see in our in our sap scenario that have the NT only increases about 1% of the right pool of EMT because people use Vehicles being more widely distributed more people willing to take the share Vehicles increases to about 27 percent. So, this keeps the average TNT occupancy above about 1.08 in the shared scenario on this because
We have the sufficient technology. We managed right Hill operations that limit the amount of density which would not be the case in the privately-owned scenarios. These changes are not only driven by Fleet operations are also driven by traveler Behavior outcome. So, one thing we can see is that the moisture changes substantially are Fleet deployments grow. So, we see trying to use girls from 6% to a high of 12 percent in some of the longer-term scenarios as households dispose a vehicle to take advantage of these vehicle fleets. However, private baby ownership, the courage's encourages additional as will be trip. So even though we have some households disposing of vehicles the level of
Travel in privately owned Autos is about 70% in the previous scenario compared to 78% of the Baseline. So, it's not too much different despite these shifts towards some of the more Transit modes and shared vehicle modes. However, this is not uniformly distributed throughout the region. So, we do find that Urban households tend to shift to Transit while Suburban shift the TNC more when they do engage in vehicle disposal. We also find that if you look even within the same scenario that households with an AV behave Much differently than households that don't so we find about an 82% or up to an 82% increase in DMT in the households that own nav on average and this is largely driven by as we mentioned previously the unloaded miles. I do to repositioning between household members as well as increased discretionary trip making so the discretionary trips tend to be about three to six miles or 30% longer on average with these additional troops concentrated heavily in the PMP because which is one many of the discretionary trips tend to take place.
As we see about people with AV spent about 30 more minutes for in travel per day that people without with that 82% VMT increase being somewhat consistent actually with some of the limited empirical studies that have been conducted two dates living a study out of the University of California Berkeley where chauffeurs were basically provided to households to mimic the effect of a Navy and they saw about an 85 percent increase in household travel under those conditions. So, we're encouraged by that finding As I mentioned previously the disease affects a Nazi uniform throughout the region so we did find in many instances that Transit and right hail can be complementary. So if we look at the change in transit mode share and this is under the high sharing scenario, I sharing how animation scenario you see that Transit mode share of far from being produced by these sheer Navy Fleet scenarios actually increases in many of the regions where Transit is already a predominant mode, so you can see sort of the Hub and spoke system in a Chicago Regional and Commuter Rail lines and then intense usage down.
In terms of increasing trends of motion are while trans not well TNC motor increases basically everywhere else. So, these two images become almost mirrors of each other. So, we see that that household as they dispose of vehicles to take advantage of the TNC mode TNT Fleet availability in some instances, especially for Transit polies are high they start to use more trans as well. Just because of outperform CNC in many of these areas. So at a high level the results across the three scenarios and for both the business as usual and program success cases are shown in this table and this can all be wrapped up in terms of development Mobility energy productivity metrics that we see in the midterm High sharing scenario with low automation levels. We have an increase in the MEP from the pair of Baseline of about 34 percent. Well that goes up to 51 to 76 percent if we have very low levels of Automation in the in the shirt fleets the largest leaves and higher
Her Nation levels that is these changes are largely driven by reductions, India. Energy and increases in vehicle speed throughout the region and then conversely if we compared to the private TV scenarios the mobility energy productivity metric only goes up about 10% So the increase there is fairly limited and that's due to the increased energy use from the additional miles traveled as well as the reduction in speed which has a Time cost. That is what a burden that applies to individual Travelers. So the high level some key findings, we find that high sharing scenarios can have a 12% reduction in GMT and energy use and if we take those High sharing scenarios and automate the vehicles that are being shared that can increase to about 18 to 23 percent while conversely if we take those automated vehicles and then make the privately owned and so the fear we have a 22 percent increase in interviews instead of reduction. So, this is driven by a lot of operational characteristics as well as behavioral changes. So, we did find that about 67 200 percent increase in transit use can be found under some of these high ridership scenarios, but if private in the private ownership scenarios households that actually do purchase of eaten
Maybe I should mention this case only about 50% of households had them in these scenarios. They those households are to purchase have about 82 percent more total VMT. So, Miles traveled day that households that do not so with that. I will say thank you. And if there are any additional questions that are not covered in the Q&A session, please feel free to contact me at the email address shown here. Thank you, Josh, lot of information there and I will say this is only a slice of what we started to learn through the players application to Chicago. Our next speaker will be Zach Needell from Lawrence Berkeley National Lab who will describe the implementation of the theme of transference of the beam agent-based model transportation system model as part of the workflow applied to San Francisco.
Zack will turn it over to you to describe the application results from being activated and think - I think difference being here. Yeah, it has been mentioned. I am going to talk about a second implication of the smart Mobility workflow the underground be model developed at Lawrence Berkeley National Lab and deployed in defense of Skopje area region in California. So, David and Josh earlier have done a great job.
Head of framing the purpose of the permeability workloads. I won't belabor my introduction, but just fine have one quick slide to give kind of a high level overview of how we're thinking about the question so you can think of a transportation system as having a supply side and the demand side. So, the supply side are the physical quantities that exist in the real world. So, travel speeds on the road Network routing on the buses empty parking spots. Demand side you have the choices that people make so how many trips people take but mode to they use what back to they take and there's a lot of Transportation modeling that focuses in on either that's by side or the pant sizes patient. But when we are talking about deep fundamental long-term changes to the transportation system the type that we're talking about with automation electrification that will take place over the course of decades. That is when this joint modeling of appliances and in the transportation system.
Really important and heard of joint supplies to me and modeling integrated now is important because it allows for detox very Supply impacts demand. So, people's choices change in response to new technology options available to them print them. But also, where the man impact Bond so different usage of this technology will be two different performance these Technologies work on them differently example of this are induced demand.
We are at widening Highway and traffic land use change and responses to change the transportation system and the introduction of new Transportation Technologies and abilities here.
And so, the integrated modeling platform in the particle’s workflow is meant to enable this integrated by the game modeling and so he's kind of a high-level picture of how the begin implementation of the workflow works. So, we start with scenario generation for defining a set of input that Define a future Transportation scenario that we want to better understand, and we end with scenario evaluation. So, taking all the outputs of this model.
And try to use it to better understand what the transfusion. Like in between remodeling policies that happen on athletes for time scales. So, the long-term year by year timescale. We want to understand our land use changes in response transportation system. We want to understand the provision of charging and refueling infrastructure and the makeup of the vehicle, please at the day by day hour by hour minute by minute time scale. We want to focus on Visions like no choice, but most people take the behavior of three
And the performances the traffic Network. So how fast part of going to the length of the road network? But then finally at the second by second scale that is where a lot of the energy interaction possible due to acceleration defense lawyers in the vehicle or potentially Communications and interactions between automated Vehicles will try to improve traffic flow. So, we want to capture kind of all these bathroom different time scales and what deformability workflow does is We
………………………….
So we're going to put a So that you know stakeholders who have them all of the developed in their region already or you know, one of the trust can plug it in Telugu framework and make the models fashion so that we can run a large number of scenarios and a lot of stakeholders to fully explore the parameters fix the potential teachers. Yeah. Thanks, if your own for your time. My email is on the slide. Feel free to reach out and find additional questions with you more about games. Thank you. Thank you, sir for running through that. It is great in thanks Josh for presentation on Polaris prior to that. We have about 15 minutes for questions and answer I will I will verbally T up the questions to both Josh and Zacks all ask you both to unmute as you answer them and you can turn on your video as well. So again, type your questions into the Q&A box and We have a question on really for both of you. The question is will first, very interesting presentations. The question is are the two packages available for download with examples. And so, I believe that means or the software packages or these models available to be downloaded and used by external folks. So maybe I will start with you Josh and give Zach chance to catch his breath.
You can apply for a week. We do have a version of Polaris that's open source and available for download the version that is used in the development of this smart Mobility workflow. And actually the holes for Mobility workflow is not currently available but we are engaging in ongoing efforts to get that deployed both as a desktop application that could be used as well as a cloud application that would be available to others but we are. Always interested and we have a government license applied to it to work with partners and get it tested. If there is interest to download the research version of this probability of workflow. So, it is available if you reach out to us that way thanks Josh and Zach, what about the availability of the model and implementation to work? Well, yeah, it is going to be model is totally open starts at the very permissive license. And so, it's the all the source code including for these Brown.
Available on GitHub. If not, we another website team that LDL back of which maybe I should have been applied. But anyone feel free to reach out to me and I can share that link but we're going to be more we all look at both the GitHub page and the overall page and yeah, so being comes out of the kind of based around the map in framework, which has a very active user base around the world can't be kind of encourage, you know. Operation and you know anyone to download the model play around with it, you know be found above your, you know have a question judgment issue on get help. We very much appreciate any kind of external users’ ignorant interest.
Great. Thanks. Um, we do have a question in the chat box rather than the Q&A box. I will remind you if that distinction is not obvious. I will please put your questions in the Q&A box, but I will read the ones in the chat box in the question is for Josh to what degree. Can you use two layers to make decisions such as no choices or new routes for transitory treats to achieve a certain. Mobility outcome expressed in MEP or incomes of missions. So again, I think the question is related to what is the predictive capability? How would you use players to make decisions? That's a great question and it's something we are actively engaged with in using Polaris for we have a few projects that are currently ongoing with Transit agencies or other industrial Partners in terms of looking at how we can more optimized. Eyes with a transit schedules Transit routes increase interaction between transit system and Rideshare systems. So we feel that with a lot of the work we've been doing on that developing the players bottle putting strong optimization capabilities in there and connecting with local partners of it would be appropriate to do it this to do this with the Polaris model and we have again hopefully ongoing projects where we would be able to deploy and
Agencies would be able to use the outcomes are. Lesions to affect a more optimal Transit System
All right. Thanks Josh. We have another question I think is probably applicable to both being in Clara's. So maybe we will start with Zach on this one. The question is what model inputs are associated with the high sharing cases. Are you just changing a couple of parameters the mode choice or vehicle ownership models, you can wear those were the modeling puts associated with the high chair in case you nailed it to Big Run the you know, there's several kind of like small.
Small differences but I hit the most fundamental one that you know that we noticed were very young. When we noticed to be more have higher leverage work higher retirement rates of the personal vehicle, please so, you know, we had a lot more household giving up their individual vehicle and into high sharing case. So, in the shared AV case we got rid of people's aversion to take me.
Shared in the Baseline case taking a Pinot a pooled right held trip look less attractive than taking them. Non poke, right he will trip because we baked in value of time aversion to spending time in hold on mobility. And we got rid of that added value concretion are in the shared automated vehicle case. I think that Josh I think that was it as well as the size of the right Sharon Fleet rights. There was an assumption from a lot of the research that has been time around maybe replacement rates. So if there were a fleet of decide if we will if we want to do assume my disposal rate on private Vehicles, it would require roughly a free to Fleet of a certain size which we said and then test it around, you know tests around the margins to make sure it was consistent with those findings, but that was the other major parameter setting. Thank you question on the three scenarios. So both of you highlighted the three scenarios that were used and applied to both Chicago and San Francisco why those three scenarios are those the only three scenarios that we can do maybe start with Josh on that we know the idea is to do a lot of different scenarios to really do, you know a full design of experiments across what we identify as all of the key parameters.
Both driving Mobility changes and associated with different future Mobility Technologies. These three scenarios were designed to get something in place as somewhat Corner cases in an initial design of experiments of extreme in terms of private 88b deployment extreme in terms of share vehicle deployment and then more of a near-term scenario which was a bit soft and from those two but these are by no means intended to be comprehensive or per fully covering the space of future mobility of this is really a starting point. And to demonstrate the workflow that we put together Zach. Yeah. Yeah, totally agree with that. These kind of represented initial Corner cases kind of kind of I think one thing at least I found that we were surprised by is maybe not surprised by this but there are a lot of knobs in the system that you know a lot of potential and but you know, and if and some of them end up being more powerful than kind of effect going in so I think you know on both sides.
We're kind of really looking forward to really exploring this more detailed parameter space, you know more thoroughly great so and along those lines so we had these three scenarios applied to two cities and smart Mobility won't .0 Chicago and San Francisco. So how generalizable are those findings, especially for San Francisco. San Francisco is might be unique among US cities so really a question for both of you, but we'll start with Zach on this one how generalizable do you think your findings are? Well, you know that that is also something that for in the process of beginning to test the past year. We started to expand the pool of cities where were deployment models with the goal of finding the center of the Venn diagram where we have implications both P being and player so that we can really
Better understand Similarities and you're going to do the models and to you know be able to better explore all through the space of potential pity, you know to get to the point where you know, even if we don't have you know, the public woman every major city in the U.S. That would be no we have had both models deployed in you know, the widest Spectrum to continue to have a better sense of how generalizable our results are.
What you know her factors there are that influence. It might be that calm. Sure. Do you have anything to add to that jobs in the context of whereas in Chicago? No, I would say that is a pretty good description of where we're at. You know, this is generalization and transferability within travel demand models is always an important topic and you know, the goal of the smart Mobility Consortium is not to understand Chicago and San Francisco for say it's understand, you know how this affects the entire country for future Mobility. So, I feel like the work were engaging with in diversifying the set of cities that were applying these models to is important and goes a long way towards demonstrating. The utility of these type of agent-based models would focus more on fundamental processes and tend to be more transferable that demonstrating the value of that within this program.
Great. Thanks. We do have a question again into the chat box about mathematical optimization. The questioning is for targeted to Zach for being and the question to ask. Is there any mathematical optimization be done indeed so I guess you could infer that several different ways? But yeah, go ahead psychic what types of mathematical optimization are you? Doing the Navy. Mm. Yeah, I have had several. I guess that's kind of fundamental ones like route choice, you know, we're you know type of a shortest path strip at routes on the transit Network Road Network in terms of kind of like more novel optimization. I would say the two biggest one we are doing a Dean is we're doing real time matching and cooling of right till trips. So, you know parent caring individual Riders together and you know matching with the vehicle did not deem we're doing it with the kind of syrup.
And based on the profile of the Maura that has got published a couple years ago, but we've made some modifications to do a lot of to run a parallel. And so yeah, so do you know that that is definitely a big optimization problem that were solving serious. So serious typically that we've kind of done tests on the trade-off between, you know, finding the best possible solution runtime and comfortable with our balance there and we also have some other optimization methods like scheduling of household all automated vehicles. To maximize the number of those trips that can be covered by a given automated vehicle and kind of going forward. There are a couple more they are going to bring it. So, you know as we start introducing Freight in the dean, there's going to be a kind of routing short traveling salesman problem that we're hoping to solve their so yeah, so we're pulling in several different parts of the literature. Are you think I would have something to add to that but I will pose a different question to you Josh one that is very timely the question has any clue yet on how COVID fears have or will affect these models that actually says these modes I will turn that to me these models how might we use our models to address the transportation disruption that has resulted. Yeah, the greed of some very timely question and something that could be addressed through these models. So, the COVID affects a lot of different impacts. It was it impacts a lot of different processes that we model from activity engagement. So certainly workers are traveling less telecommuting more people are engaging in less socializing activities recreational activities and so on as well as if we focus specifically on modes it's that was the intent there is a strong impact that we've noticed in working with our partners and
CTA the Chicago Transit Authority and others in terms of overall Transit usage TNC uses all of these things are dropping, you know, as people avoid these type of modes due to either fear or social distancing requirements and so on. So certainly, these are effects that are being moving in the real world and that can be simulated with within each model. Okay. Thanks Josie. You have anything to add to that hopefully will be at the point to be able to share something publicly at some point. But I think you know this kind of joint integrated digital modeling is especially important for situations where we're modeling big changes the transportation system, you know, we need to integrate and I think COVID is a prime example of you know any Leah we're taking a
Kind of simplified model and just take the parameters might not work very well because just there's so many fundamental parts changes happening. And so, I think it's definitely these kinds of models are like really well suited for you know modeling these kind of questions would be something that you know is today someone's got to make changes. It is hard to validate everything perfectly, but I think they're really good at being able to you know map out the kitchen with competitions beautiful way. Yeah, absolutely. I think it is an example of the capabilities of the tools that you've built the labs that we are able to now apply them to some of the questions around. What do what do a return to normal look like in terms of transportation? We have another question in on predicting future scenarios. Are you currently doing any studies on how to predict realistic scenarios on modeling the response?
Technologies that do not exist right now from my fav fully automated Vehicles do not exist right now, but there's sort of a Glide path of road map to get there. What about other technologies that don't exist right? Now? How do we model user’s response to those Josh? Maybe I will start with you on that. Yeah. That is always an interesting behavioral research challenge, right understanding how people would respond to things which don't actually exist a lot of that. Work is done mostly or USA. Usually through stated intentions or stated respond response surveys where you just ask people. Hey, what would you do if something like this exists which has its own set of challenges and problems in terms of representativeness? We also attempt to learn by analogy, right? So, we don't necessarily know exactly how people would say spend their time in an automated vehicle, but there are things we can learn from how people spend their time in Uber vehicle. It is not exactly
Same arms are right your vehicle's not exactly the same but there are Analogies you can draw between those and same with the how people spend their time and Transit and that's just one example for a value of time changes in an automated vehicle for the talking about more technology more broadly. So things we could deploy with connectivity or looking at how people would react to say no heads of messages or messages from infrastructure through Eco approach to departure at an intersection or so on a lot of these things are actually being tested in the field and we've tried to partner with The industrial Partners or other research agencies that are looking them as does available. But yeah, the range of future Mobility Technologies is Broad and it's very difficult in many cases to understand what those sexual impacts would be but we do we do the best we can with what's available great. Thank you, Josh., We are just a minute past the hour. So, I am going to end the Q&A there again. Thank you, Josh Auld from Argonne National Lab and Zack Needell from Lawrence Berkeley National Lab for Presenting the work that you have been doing over the past number of years. I will remind you all that again. Just as number three in a series of webinars on the screen. Now you see that there are six Capstone reports that are available for download at the link is provided one for each pillar of the first phase and smart Mobility as well as one covering the implementation and results from workflow. So, we invite you to visit the link download the reports and all you are at that link you can also
The next slide. There we go. You can also register for upcoming webinars, which are listed there on your screen. You can also download the presentations the transcripts in the video from the webinars that we have already had. So again, thank you everyone for your time today and for joining us. Hope you found this information informative if we look forward to talking with you again at on webinar number 4 on November 5th where we will dive into some of the research that the
Should do on the benefits of connected and automated Vehicles. Thanks again. Stay safe and have a great day.