<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Yuso blog articles (in English) - Medium]]></title>
        <description><![CDATA[Find here all our articles about urban transportation, mobility or dispatch - Medium]]></description>
        <link>https://medium.com/yuso-articles?source=rss----d26e8c06f568---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>Yuso blog articles (in English) - Medium</title>
            <link>https://medium.com/yuso-articles?source=rss----d26e8c06f568---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 20:00:22 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/yuso-articles" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The puzzle of dispatch, part 2 : Ride-pooling]]></title>
            <link>https://medium.com/yuso-articles/the-puzzle-of-dispatch-part-2-ride-pooling-1136cc051ea1?source=rss----d26e8c06f568---4</link>
            <guid isPermaLink="false">https://medium.com/p/1136cc051ea1</guid>
            <category><![CDATA[ridesharing]]></category>
            <category><![CDATA[optimization-algorithms]]></category>
            <category><![CDATA[algorithms]]></category>
            <category><![CDATA[dispatching]]></category>
            <dc:creator><![CDATA[Yuso]]></dc:creator>
            <pubDate>Wed, 05 Sep 2018 12:24:49 GMT</pubDate>
            <atom:updated>2018-09-05T12:23:56.053Z</atom:updated>
            <content:encoded><![CDATA[<p>After analyzing in<a href="https://medium.com/yuso-articles/the-puzzle-of-dispatch-part-1-advance-booking-455893821fa"> a first article</a> the best way to dispatch in advance bookings, we are now going to study the case of ride pooling between several passengers. As a matter of a fact, ride sharing is becoming an increasingly important issue in transport and logistics industries.</p><p>From a fleets’ point of view, ride sharing allows to optimize resources and to make their business more profitable. Creating an optimized route and simultaneously transporting several packages or several passengers in the same vehicle (like it’s done for example by Via, or UberPOOL) enables to minimize the time spent empty and the driver’s working time, in a sector that is highly sensitive to operating margins.</p><p>From clients’ point of view, the use of a shared transport service between several passengers (or between several packages in the case of delivery) makes it possible to obtain more attractive prices, in exchange for a slightly downgraded experience (longer waiting time, detours during the ride, etc). For instance, the fares proposed by UberPOOL are 25% to 30% cheaper than the price for an identical ride using UberX.</p><p>Finally, from cities’ point of view, ride sharing helps relieving traffic congestion, and reshape current individual means of transport (taxis or personal cars) as a new form of shared transport. As mentioned in our previous article, an MIT study based on New York taxis reservation data shows that is it possible to share up to 97% of the requests of rides, without significantly impacting the passengers’ rides ( the time added to the journey remains less than 3 minutes).</p><p>The taxi is getting closer and closer to become a bus. For example, in medium-sized or rural towns, there is an increasing demand for school based transport or the transport of PRMs (People with Reduced Mobility). Another example would be path pooling, it could be used to transform public transport into a smart and digitized model, which would be not based on rigid lines at fixed timing but rather on dynamic routes, that adapt according to demand.</p><p>If we now consider the subject of sharing a journey from the point of view of dispatch, it raises three major issues, between which we must find a balance : a technological issue, an operational issue, and finally an issue in terms of user experience.</p><h4>Pooling rides from a technological point of view</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*knSD36XwuRrqbu3zgB965w.jpeg" /></figure><p><em>“This is the story of a travelling salesman”</em></p><p>The technological challenge associated with pooling is actually close to a well-known problem in computer science and optimization, known as the “travelling salesman problem” (TSP).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/974/1*HfZC8Xw5lf-BZ7Vo1QDcxA.jpeg" /></figure><p>This problem has a very simple statement: a travelling salesman must go successfully to several cities while minimizing the total distance he travels. The distance between each of the cities is known, so it is necessary to determine in what order the salesman should visit the cities.</p><p>If the problem statement seems easy, it is actually more complex to solve, because it requires to test all the possible combinations: if the traveler has to go to 10 cities, it is 10! different paths to calculate, i.e. 3,628,800 combinations. If he has to go to 20 cities, then you would have to calculate 20! possible paths, which represents more than 2 billion billion solutions! Even with a million calculations per second, it would take more than 77,000 years to test everything.</p><p>The issue here is not so much to solve the problem, which remains algorithmically relatively simple, but to solve it in a very limited time. In fact, a fleet operator should be able to provide in only a few seconds an answer to a customer, who would like to for example to book a ride that starts immediately.</p><p><em>The case of an immediate departure</em></p><p>When a ride with an immediate departure must be pooled with other rides already planned, we must take into consideration as vehicles able to pick up a client all the vehicles circulating, even if another passenger is already on board. In this case, the objective is not to minimize approach time, but to find the most optimal route. For example, a vehicle could realize an intermediate stop before picking up the new passenger.</p><p>In terms of calculations, we are here faced with problems very similar to those of advance reservation, with a large number of possibilities in the choice of the vehicle. The only possible filter at the vehicle level is based on the time windows for the pick-up or the drop-off of the passenger, which will limit the possibilities of sequencing the rides.</p><p><em>The case of an advance booking</em></p><p>When it comes to advance booking, the problem is even more complex to solve: the response speed to customer is not as important in this case, but the number of calculations is very important to establish the schedules of the drivers, because it is essential to calculate all intermediary travel times between stops. The diagram below allows to better visualize the arrangement of different rides.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RA7PzII1srF9HV2mTO3jWA.jpeg" /></figure><p>Let’s try to determine the number of calculations needed with this new configuration. Consider, for example, a case where there are 10 available vehicles, and they receive 100 requests for bookings, to be divided among these 10 vehicles. We will try to calculate the number of intermediate travel times. For this, we must consider :</p><ul><li>200 intermediate stops (one point to pick up and one point to drop off each of the 100 journeys)</li><li>10 starting points of vehicles (1 stop per vehicle)</li><li>10 points of arrivals of vehicles (1 stop per vehicle)</li></ul><p>That is in total 220 points, from which it is necessary to calculate the routes to the other 219 points. Therefore, there are 220 * 219 = 48 180 travel times to calculate.</p><p>If x is the number of vehicles and y is the number of journeys, the number of intermediate travel durations N is obtained with the following formula :</p><p>N = (2x + 2y) *(2x +2y-1)</p><p>In this case, the number of computations is quadratic to the entry n. In computer science, it’s called a complexity in O(n²). Calculating travel times in a limited time remains possible.</p><p>However, our work does not stop there: once these journey times are calculated, it is still necessary to determine which is is the best schedule arrangement. And this is where the problem shows up, because the number of possibilities of planning is factorial to n. It’s called in this case a complexity in O(n!). If we take again the example aforementioned, then we would have (200+10+10)! = 220! possible arrangements, which is more than the number of atoms in the universe! It is therefore impossible to calculate in this way the most optimal schedule, whereas 100 requests divided over 10 vehicles actually corresponds to a low volume of passengers.</p><p><em>Using a local map</em></p><p>In order to calculate these intermediate travel times, most dispatch solutions are based on data provided by Google via its Google Maps API. However, the number of requests per second is limited, and the service becomes paying beyond a certain number of requests. This issue limits the number of possible calculations and can ultimately cost a lot, especially since Google has just sharply increased its prices.</p><p>To not be dependent on a third-party service and to be able to make an unlimited number of requests, it may be wise to develop one’s own mapping solution. In this scenario, it must be possible to guarantee a very short response time, especially in the case of a reservation with an immediate departure. This is the choice we have made at Yuso. We have developed our own local map that allows intensive calculations, or in other words, compute a very large number of travel times in a few seconds.</p><p>An example of a local map is available down below. In this map, the roads are all prioritized according to the volume of vehicles passing through. Thus, the structuring axes are in red, the second level axes are in orange, etc. In this map of Paris, provided by OpenStreetMap, the ring road is in red because it is an axis that allows you to go around Paris very quickly.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*Dp8rLMtyTvfN0f5e" /></figure><p>As mentioned in the article “<a href="http://algo2.iti.kit.edu/schultes/hwy/hhJournalSubmit.pdf">Engineering highway hierarchies</a>”, in order to travel as quickly as possible, the objective is to reach a major axis as soon as possible, and then go down the hierarchy of axes when approaching the arrival point. Through this way of building a route, which is actually quite intuitive for a driver, can drastically reduce the number of calculations necessary to establish the number of travel durations.</p><p>From a mathematical point of view, this map is actually a graph that models Paris, through edges (streets) and nodes (street corners). A travel time is assigned to each edge, similarly to the travelling salesman problem. This simplifies the mapping of a city and thus simplifies the calculations of travel times.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Hm03SU2TTG4zXmxvXMpYoA.png" /></figure><p>The use of this kind of graph with hierarchy of axes makes it possible to reduce the computation time up to 75% compared to a graph without hierarchy, and to also improve the precision of the results, because its mesh is denser. The only drawback of this solution is that it requires to create a similar build for every city where a company wants to deploy a fleet, which can be a very time-consuming process depending on its scope and the number of axes to prioritize.</p><p>If then we plan on to optimize the schedules for drivers, we must test all possible combinations between the different travels, and in the case of shared rides, this issue increases substantially the number of calculations. To minimize the number of calculations, it is necessary to pre-filter the vehicles (vehicles located too far away, or ones who have already many passengers on board, etc).</p><h4>Pooling rides in operational terms</h4><p>The fact of wanting to mutualize several rides may seem like an attractive idea at first sight, when we look at it from an operational point of view, because we will be able to make more trips with fewer vehicles. In reality, this type of service rigidified and controlled in order to be really effective and interesting from an operational point of view.</p><p>As a matter of a fact, a transport service offering ride pooling between several passengers cannot work if we do not put some constraints (in terms of availability, authorized routes or zones), because then the probability of pooling rides can strongly decrease. For example, a journey in the off-hours from London to far southwest suburbs will be unlikely to be shared with a similar journey, while a trip to Paddington during rush hours in the morning will be more likely to be shared with similar requests for a booking.</p><p>Here are some examples of constraints that can be used to maximize the probability of pooling multiple trips:</p><ul><li>Define as points of pick-up or drop off some strategically located virtual stops, to which the passenger will have to go to (rather than being picked up or dropped off at the exact address he requested). This is what was put in place by Via in New York or Washington or by Bridj in Sydney for example.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*c1YubNuq3HnIpwL4" /></figure><ul><li>Define hourly constraints regarding the availability of the service (for example, only available during the rush hour in the morning and evening, to serve commuters).</li><li>Authorize some areas or pick up corridors as “hot areas” that have a lot of demand ( this model is used by Chariot in San Francisco)</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*0fzlGwpHRveRNchi" /></figure><ul><li>Make the client wait for a few minutes before offering a ride, to maximize the chances of finding a similar route in a way that the ride can be pooled.</li></ul><p>On the other hand, if the scheduling or the operational constraints laid out are too restrictive (service is only available during a reduced time period, or in a limited geographical area), the service will not work either because the probability that a trip will meet exactly the criterias requested is very low.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*ZCBRt-ABFyzFnFzd" /></figure><p>The difficulty here is to find the right balance between the constraints; they must ensure profitability to the service and a minimum occupancy rate, but they do not have to frustrate the client and decrease the quality of their user experience when using the service.</p><h4>Last, but not least: the user experience</h4><p>Even if a customer who uses a shared mode of transport rather than an individual mode of transport expecting to have a less “optimal” trip, it should not be a disappointing experience. It seems unwise, for example, to share a ride between passengers who have really distant arrival addresses, or to multiply their travel time by three, even if they pay a reduced price compared to a non-shared mode of transport.</p><p>The case of UberPOOL service is very interesting regarding this issue: Uber initially chose to offer door-to-door service, and picked up customers from the addresses they requested. This model, which seems to be the most satisfying for the customer, is actually not the most optimal. Picking up or dropping off other customers at specific addresses forces the vehicle to make detours in narrow streets (and in Paris, there are <em>a lot</em> of one way streets) which leads to wasting a lot of time. Following the claims of its users, Uber has decided to change their model and offer another one called “<a href="https://www.theverge.com/2018/2/21/17020484/uber-express-pool-launch-cities">UberPOOL Express</a>”, tested in the United States. It requires its customers to walk to meeting points ideally located on major roads to minimize the loss of time associated with picking up or dropping off customers.</p><p>Hence, this implies that the customer must walk for a few minutes to get to a point of pick-up, but the overall benefit is important : the other passengers save time individually on their journey, and the driver can be available for more rides and have a better hourly wage.</p><p>This is also the conclusion made by Via in New York, which requires its users to go to the points of pick-up that are located at street corners. They also must wait for a few minutes before getting a proposal of a journey, and confirm the reservation within 30 seconds, because the place is pre-booked in the vehicle.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*3Qx_rsouf5QD2k55" /></figure><p>Just as when a shared mobility service places operational constraints, the challenge here is to balance the effort required by the user and the overall quality of the service provided. It is the balance between the attractive prices and optimal operational constraints that increases the efficiency and comfort of the passenger. This will be the challenge for all the services of bus on demand, whether they are private or public, in the following years.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1136cc051ea1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/yuso-articles/the-puzzle-of-dispatch-part-2-ride-pooling-1136cc051ea1">The puzzle of dispatch, part 2 : Ride-pooling</a> was originally published in <a href="https://medium.com/yuso-articles">Yuso blog articles (in English)</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What kind of future for urban mobility?]]></title>
            <link>https://medium.com/yuso-articles/what-kind-of-future-for-urban-mobility-b5e14d944a6b?source=rss----d26e8c06f568---4</link>
            <guid isPermaLink="false">https://medium.com/p/b5e14d944a6b</guid>
            <category><![CDATA[autonomous-cars]]></category>
            <category><![CDATA[transportation]]></category>
            <category><![CDATA[urban-mobility]]></category>
            <category><![CDATA[dispatches]]></category>
            <category><![CDATA[on-demand-transit]]></category>
            <dc:creator><![CDATA[Yuso]]></dc:creator>
            <pubDate>Thu, 21 Jun 2018 15:27:46 GMT</pubDate>
            <atom:updated>2018-07-06T08:46:37.469Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V8XKz2JuDOyuVi5N7GBmkA.jpeg" /></figure><p>These last few years have seen the rise of numerous deep and long-lasting changes in the mobility sector. The urban mobility paradigm drawn in the sixties with the widening access to private cars is fading, opening the door to new trends.</p><p>At Yuso, as mobility specialists, we are convinced that the revolution experienced by urban mobility in the next few years will be based on three pillars: the car of tomorrow will be first electric, then shared and finally autonomous. These three aspects are in fact closely linked. The growing success of electric cars will be a great springboard for the development of on-demand shared transportation, and eventually for autonomous car, as we will show in this article.</p><h4>Cars (and cities) of tomorrow will be green</h4><p>We can notice a global trend in the last few months: the rise of electric or hybrid cars. This trend is mainly pulled by governing bodies and by municipalities. They are very aware of the environmental and sanitary issues linked to the use of cars (air pollution, noise pollution, global warming, oil scarcity), and they are more and more willing to ban from cities cars with diesel or gas engine.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*Fdk1U_yvG0UvlscbEwGHAA.jpeg" /></figure><p>C40 Cities, a non-profit organisation that brings together the 40 largest cities in the world set as a priority the eventual extinction of diesel vehicles. Paris and Mexico, which are trialing a new program, <a href="http://c40-production-images.s3.amazonaws.com/other_uploads/images/1011_Mayors_Air_Quality_Commitment_Jan17.original.pdf?1485784551">committed into making them disappear by 2025</a>.</p><p>If the electric cars were first limited to a niche market, all car makers are now taking this path. To be able to adapt their models to changes required by cities, all car makers are rushing to extend their electric range, and replace their traditional models. Their objective is to lower costs and to optimize production processes to turn electric vehicles into mass-market products, especially thanks to technical improvements of batteries. This was highlighted by Carlos Ghosn, CEO of Renault, when he revealed in October 2017 the <a href="https://www.alliance-2022.com/blog/drive-future2017-2022-renaults-new-strategic-plan/">strategy of Renault-Nissan-Mitsubishi group</a> to improve the profitability of electric vehicles.</p><p>If electric vehicles are for the moment more expensive than traditional vehicles, they have nevertheless a longer life cycle and lower maintenance costs. Indeed, contrary to cars equipped with combustion engine, which can experience an engine or clutch breakdown, the only weakness of electric vehicles is their battery, which needs to be changed regularly.</p><p>However, the electric cars’ advent will not be a magic bullet for municipalities. The main challenge for cities, beyond the pollution problem, is also to fight against streets saturation and traffic jams. The best way to go, in addition to the expansion of public transport networks, is to call into question the individual car model. Indeed, an individual car is actually occupied by 1.2 passenger on average for commuting rides (home to work rides), and this has a huge impact on cities’ congestion. Moreover, a car is driven only during 5% of its lifetime, and spends the most of it parked, using a lot of square meters in urban landscape.</p><p>Leaving the individual car and switching to shared vehicles, or on-demand vehicles, to optimize their use seems to be a logical and a good option. It makes even more sense because these new electrics cars will be able to be driven during a longer time, and on a longer distance.</p><h4>Sharing vehicles for better cities</h4><p>In the last few years, people have shown less and less interest in individual cars, and the market is evolving through a <em>MaaS</em> (“Mobility as a Service”) model, based on shared vehicles available on-demand. The general expression “Mobility as a Service” designates in fact different car-sharing services.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w3EXfyQCB7Im3XvMmpOOhA.jpeg" /></figure><p><em>On-demand vehicles</em></p><p>Consumers are more and more numerous to choose peer-to-peer services to rent a car, like <a href="https://www.drivy.co.uk/">Drivy</a>. A good way for owners to make profitable an unused vehicle, and a good way for renters to not be burdened by a car they almost never use.</p><p>In the last two years, the number private services offering on-demand vehicles have skyrocketed: <a href="https://www.car2go.com/US/en/">Car2Go</a> for cars, <a href="https://joincoup.com/en/berlin">Coup</a> and <a href="http://www.cityscoot.eu/paris/?lang=en">Cityscoot</a> for electric scooters, <a href="https://mobike.com/">Mobike</a>, <a href="https://www.ofo.com/us/en">Ofo</a> and others for bikes. Finally, using taxi or minicabs for occasional rides is a strong trend on the urban transportation market. For instance, 34% of <a href="https://www.lyft.com/">Lyft</a>’s users <a href="https://take.lyft.com/economic-impact/Lyft-Drives-Economy.pdf">have declared in a survey</a> that they didn’t buy a car because they prefer using the service for occasional urban rides.</p><p>We can consider all these elements as evidences showing the progressive disfavor for personal cars. Nevertheless, they need to be tempered: if minicabs and taxis are great alternatives to individual cars in urban areas, personal cars remain the queens in semi-urban or rural areas. It makes economically more sense to use one’s personal car rather than a taxi or a minicab in case of intensive use. If sales are decreasing little by little, several million cars are still sold every year.</p><p><em>Urban car-pooling</em></p><p>The second model of car-sharing is urban car-pooling. As of now, apart from local initiatives, there isn’t any service which has been able to settle its position on the market at a huge scale and on the long term, like Blablacar has succeeded on the long-distance car pooling for instance. Some tests have been made at local scale, especially in North America, but none of them has really met broad success until now.</p><p>Whatever the city, the different trials that have been conducted are all based on an individual car model, where a driver agrees to pickup passengers during his ride. Some obstacles are slowing down those initiatives, especially geographical and time constrains: to avoid detours and increasing travel times, drivers must take passengers whose route is really similar. Moreover, contrary to long-distance car-sharing, the financial incentive -not more than a few cents- is really low. Finally, urban car-sharing needs a critical mass (an important offer and demand, on a constrained territory) and a deep change in minds, while the car is perceived as a private and personal space by many people.</p><p><em>Ride-sharing services</em></p><p>This is the most advanced model of ride-sharing, in which the car doesn’t belong to an individual who can rent it or share it temporarily, but to private or public organizations, which manage and regulate ride-sharing between different passengers. This is the real model of <em>shared</em> vehicle, with the end of car as an individual property and the beginning of ride-pooling, making this type of service quite close to a public service.</p><p>In this perspective, we can imagine that on very dense urban territories, with numerous ride bookings (for instance Manhattan district or Paris city), the taxi rides can be shared automatically between several passengers. A study conducted by the MIT in New York has shown that it’s in fact possible to pool 97% of taxi rides, with a minimal impact on passengers (the increase of ride length is under 3 minutes). By doing so, it would be possible to cover the current demand with only 3 000 vehicles, while they are 13 000 today.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4tkQl5fLx8uvSYJS2CMykA.jpeg" /></figure><p>We can see here the first sight of tomorrow’s public transit, based on on-demand buses. That’s why Yuso team is working on the development of a real-time dispatch and optimization tool, able to pool rides between several passengers in the same vehicle. This tool can be used by taxis, private hire fleets, or even public transit services. For instance Via launched in New York <a href="https://ridewithvia.com/2017/06/introducing-shared-taxi-new-way-ride/">a partnership</a> with the taxi booking app Curb.</p><h4>Towards the automation of cars</h4><p>Sharing a vehicle between several passengers, as described previously, is a good way to improve the average occupancy rate, and will have a positive impact on booking price (you can read <a href="https://medium.com/yuso-articles/the-puzzle-of-dispatch-part-1-advance-booking-455893821fa">our previous blog article</a> dealing with this subject). When you take a taxi or a minicab, le cost corresponding to driver salary (gross revenues minus vehicle’s maintenance costs, oil, insurance, etc) is equal to 50% of total cost, according to estimations made in France. From an economical point of view, shifting to transportation services based mainly on autonomous cars seems the best option in the long run.</p><p>This last evolution is for sure the most ambitious and the longest one. It will take several decades before this new model becomes universal and available at a large scale. With this in mind, numerous mobility and tech actors are currently working hard on autonomous cars projects.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DswdM5NEUTAE0--ITH5Cfg.jpeg" /></figure><p>The likely democratization of autonomous vehicles rises a question: will consumers who can afford such vehicles use them even more? For car models presenting the higher degree of autonomy (class 5), and which don’t need any driver, we can imagine that the car will be an extension of home or office, where passengers can do their daily tasks.</p><p>Beyond the impact on individual cars, the revolution of autonomous vehicles might impact before all public transit services: better and more flexible services, extended hours, denser networks,… These changes will be really beneficial for urban, sub-urban areas, and even for rural areas which are missing proper transportation offer for the moment. Public transit based on autonomous vehicles might make accessible areas where there was not alternative to personal cars.</p><p>Some trials have been launched in 2017, like for instance the experiment made by Renault and Transdev in Rouen Technopôle with autonomous Zoé cars.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/0*stbnzgoMIMK2hQg3.jpg" /></figure><p>In the same way, trials are more and more numerous in France, especially in Paris. Some tests are currently taking place in La Defense (major business center in Paris) thanks to <a href="https://navya.tech/en/">Navya</a> vehicles operated by Keolis, or between Austerlitz and Lyon train stations, with <a href="http://www.easymile.com/">EasyMile</a> autonomous shuttles operated by RATPDev.</p><p>These experiments are made for the moment in limited areas with few traffic, on dedicated roads, with a supervisor on board to secure passengers. In the long run, shuttles may be completely autonomous, but will need a specific symbol or signal to warn drivers.</p><p>In this context, a company like <a href="http://www.yuso.tech/">Yuso</a> can play a key role, as a developer of automated dispatch and ride optimization solution.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b5e14d944a6b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/yuso-articles/what-kind-of-future-for-urban-mobility-b5e14d944a6b">What kind of future for urban mobility?</a> was originally published in <a href="https://medium.com/yuso-articles">Yuso blog articles (in English)</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[2017 Review and 2018 perspectives for Yuso]]></title>
            <link>https://medium.com/yuso-articles/2017-review-and-2018-perspectives-for-yuso-859abec6d721?source=rss----d26e8c06f568---4</link>
            <guid isPermaLink="false">https://medium.com/p/859abec6d721</guid>
            <category><![CDATA[dispatches]]></category>
            <category><![CDATA[product]]></category>
            <category><![CDATA[taxi-booking-software]]></category>
            <category><![CDATA[software]]></category>
            <category><![CDATA[delivery]]></category>
            <dc:creator><![CDATA[Yuso]]></dc:creator>
            <pubDate>Thu, 21 Jun 2018 13:59:36 GMT</pubDate>
            <atom:updated>2018-02-08T17:29:23.105Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>2017, a year full of success and changes for Yuso</strong></p><p>2017 was a really exciting year for Yuso, and we reached a real turning point, both from a business and product point of view.</p><p>It was the second year we marketed our dispatch management solution, and we proved its reliability, even for large fleets with hundreds of vehicles. In one year we went from 30 to 50 clients, multiplying by two our revenue, and by four the volume of rides dispatched on our platform. In February 2017, we successfully deployed our solution for largest taxi company in Luxembourg WEBTAXI, whom was compared to Uber by the Luxembourgish Prime Minister, Xavier Bettel.</p><h3>Silicon Luxembourg on Twitter</h3><p>LID2017: &quot;We don&#39;t have Uber here in Luxembourg, we&#39;ve Webtaxi! We need to balance between regulation and innovation&quot; - Xavier Bettel</p><p>We also get closer and closer to our global vision and final objective, which is to offer an all-in-one platform, dedicated to any kind of on-demand mobility service. We started in 2017 by developing some features dedicated to last-mile logistics, and by testing them with some of our clients.</p><p>Last but not least, we are backed since August 2017 by the world leading car-makers, Renault-Nissan group, through its subsidiary RCI Bank and Services. This acquisition has allowed us to have a better financial visibility, and to reinforce our team by hiring 7 talented people, who bring new skills and expertise to the different teams (backend development, UX design, front end development, R&amp;D, marketing).</p><p>Regarding product enhancement and features, we put in 2017 a high focus on improving our existing solution dedicated to e-hailing (taxis and private hire fleets), but also on developing a new version of our dispatch management solution, adapted to delivery services and last-mile logistics.</p><p>Here is a summary of the main features released in 2017:</p><p><strong><em>E-hailing improvements</em></strong></p><ul><li>Dispatch adaptation for large fleets, with revenue balance and multi-criteria algorithms</li><li>Peak-hour management to reduce delays and optimize occupancy rate</li><li>Operational productivity with UX improvements to ensure better action fluidity</li><li>Creation of external invoicing mode</li></ul><p><strong><em>Developing Delivery DMS</em></strong></p><p>This solution is based on similar technology and products than e-hailing solution, with additional products and features:</p><ul><li>New dispatch algorithms that enable orders pooling depending on multi-criteria (package size, vehicle capacity,…)</li><li>New back office interface: booking page, dashboard, driver app and web booker</li></ul><p><strong><em>Interfaces refactoring</em></strong></p><p><em>New web booker</em></p><p>The customer web booking interface was entirely refactored and rebuild from a UX/UI point of view.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*T9oIsJ87I_-XnL_lr1aPNQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zlx9RjEeaae5fuCRBd_lUg.jpeg" /></figure><p><em>New customer app</em></p><p>We create a whole new experience for your customers, with a new UI and UX design for iOS and Android app.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fgkumAUaw5ktiLhCBhJ6Og.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hrca0HdnWxsTGKnkFHgOEg.jpeg" /></figure><p><em>New driver app</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pKNZEHR0L0_S9alwxKPpVQ.jpeg" /></figure><p><strong><em>Third-party services integration</em></strong></p><p>We worked a lot in 2017 on integrating third-party services. This allows our clients to use complementary solutions and services to increase productivity and efficiency is their daily operations. Here are a few examples of the services we integrated:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LgcKC4rcy29uA-kGJwKQ0A.jpeg" /></figure><p><strong>Our objectives for 2018: hold steady and go faster</strong></p><p>2018 will be really a decisive year for the team, with many challenges to take up and upcoming novelties. The three key words for this new year will be:</p><p><strong><em>Innovation</em><br></strong>Our main objective is to keep diversifying our products and being able to embrace future of mobility. The biggest challenge will be to developed a whole new dispatch solution, tailored for ride-sharing services. Our ambition is to come up with the world first all-in-one platform, able to manage and combine different types of on-demand mobility services.</p><p><strong><em>Scalability</em><br></strong>We want to put a high focus on making our DMS tools more sturdy and able to scale with growth of our clients activity. We already expect to multiply by ten the number of rides dispatched thanks to our solution in 2018. The scalability will mainly focus on process automation concerning drivers and operational management, as well as creation of a simulation tool to make better analysis and understanding of demand flows.</p><p><strong><em>User Experience<br></em></strong>We will start a complete refactoring of our product interfaces (driver app, dashboard, ride-sharing customer app), to offer the best experience possible to your customers or your drivers.</p><p>As a conclusion, the new features for 2018 will be oriented both on our historical solution dedicated to fleet management, but will also focus more and more on features linked to ride-sharing and delivery business. The objective here is to transform our solution into a global multi-service platform, that enables fleet to manage several mobility services at the same time: taxi and delivery for instance, or taxi and ride-sharing. The diversification of mobility services offered by a fleet will be a really key aspect of <em>MaaS </em>(Mobility as a Service), and we will put all our efforts to inject this vision in our solution.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=859abec6d721" width="1" height="1" alt=""><hr><p><a href="https://medium.com/yuso-articles/2017-review-and-2018-perspectives-for-yuso-859abec6d721">2017 Review and 2018 perspectives for Yuso</a> was originally published in <a href="https://medium.com/yuso-articles">Yuso blog articles (in English)</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The puzzle of dispatch, part 1: advance booking]]></title>
            <link>https://medium.com/yuso-articles/the-puzzle-of-dispatch-part-1-advance-booking-455893821fa?source=rss----d26e8c06f568---4</link>
            <guid isPermaLink="false">https://medium.com/p/455893821fa</guid>
            <category><![CDATA[dispatching]]></category>
            <category><![CDATA[transportation]]></category>
            <category><![CDATA[taxi]]></category>
            <category><![CDATA[research]]></category>
            <category><![CDATA[algorithms]]></category>
            <dc:creator><![CDATA[Yuso]]></dc:creator>
            <pubDate>Thu, 21 Jun 2018 13:58:39 GMT</pubDate>
            <atom:updated>2018-05-04T13:42:45.330Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kGFZNkIY3Kc2suclqwrh8A.jpeg" /></figure><p>In this article, we are going to tackle a complex subject that means a lot to the Yuso team: dispatch. Dispatch, in the sector of mobility, is the allocation of a ride to a vehicle, while optimizing the process to identify the best vehicle. These are considered as key elements in the field of transport and logistics.</p><h4><strong>The occupancy rate challenge</strong></h4><p>The reason why dispatch is a key element is its direct impact on the occupancy rate of a vehicle. In fact, the occupancy rate is a key performance indicator that has a great influence on the rides’ price, and on their profitability. Whether they drive taxis, minicabs or PHVs (Private hire vehicles), drivers want to maximise the occupancy rate to avoid driving without passenger, and thus without making money. The advent of new technologies and the digitalization of the mobility sector have allowed to improve the occupancy rate significantly.</p><p>In some cases, Uber drivers can have up to 60% as an occupancy rate, against 45% only for taxis. This remarkable difference has a direct impact on the price paid by the client. Optimization of the occupancy rate for Uber’s drivers is not only linked to the attractive prices they proposed (which increase the demand), but also to technological progress: the ability to geolocate of a driver through a smartphone in real time and the calculation of the estimated ride time while integrating real time traffic (via Google Maps), allow a better rides dispatch and a better assignment of vehicles .</p><p>To have a better understanding of the word “dispatch”, we have to consider that it actually refers to various types of computing algorithms, able to solve problems with different scales of complexity. For instance, those algorithms allow to minimize the number of kilometers driven or the time spent vacant, and to operate a transportation service in a way that is economically sustainable.</p><h4><strong>Advance bookings management</strong></h4><p>When looking at the on-demand transportation, we are able to distinguish between two models of bookings (which can sometimes coexist within the same service): immediate booking and advance booking. If the first model is relatively an easy case to treat, from an algorithmic perspective, the second model is more complex to compute as well as to operate, and can be a real puzzle.</p><p>As a matter of a fact, if the promise to accept advance bookings seems like an enticing proposition for the customer, it becomes a strong engagement and a real constraint once it is accepted by the transporter. Regardless of when the ride is scheduled and its duration, the transporter must find a driver and deal with uncertainty to ensure the transportation of his customer. It is not by chance that for a long time, Uber refused to offer its customers a service that enables advance bookings. Advance bookings is a real dilemma that Yuso on which is developing solutions of intelligent dispatch.</p><p><strong><em>Real time assignment</em></strong></p><p>The first method to manage advance bookings is to consider them as immediate rides and therefore, to assign them to a vehicle a few minutes before the scheduled pick up time. This is how fleets operate with a high density of demand. The operational problem with the this solution is twofold. First of all, to make sure that the driver will be on time, the booking has to be allocated in advance, typically between 10 and 40 minutes before the beginning of the ride, forcing the driver to wait for his passenger. Second of all, in this scenario, the fleet’s capacity is not considered, as well as the fact that a driver can be available “randomly” in the right place at the right moment.</p><p>In algorithmic terms, we are here in a relatively simple case where the ride assignment to a vehicle is based on the shortest ETA, and where it is also possible to introduce additional optimization criterias (for example : revenue’s balance between drivers, or a priority order in a rank zone). We can go a little further in the optimization criterias by introducing demand forecasts by time and by geographical area. This slightly complicates calculations but remains a relatively simple problem to solve.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*qhCUfpLOwfXJvYu4VI4TXg.gif" /></figure><p>This is the dispatch logic mainly used by market leaders, which have a large number of drivers and high demand, such as Uber or Chauffeur Privé in Paris for example.</p><p><strong><em>Advance assignment</em></strong></p><p>On the contrary, in the case of limited availability of drivers, it is a risk to accept all booking requests, and treat them as immediate rides, because there is a chance that no driver will be available at the chosen time near the pick up address. In this case, the transporter will prefer to assign in advance the booking to a driver, to be certain it will be ensured. Advance bookings are thus assigned in priority over immediate rides, even if they are sometimes less interesting from an economic point of view (more miles driven unoccupied).</p><p>For example, a trip at 9am on a weekday, assigned in advance at the far end of London’s suburb, will require a significant approach time from the driver, because it is unlikely that he would be nearby, and thus the driver will make a few pounds. The time/income ratio is then not interesting. It would have been more economically interesting for the driver to place himself at that time near an airport for example, where the demand is strong and the price of the rides is high.</p><p>Therefore, we have developed at Yuso a tool able to manage ride assignment based on planning construction. This enables to automate this process, that is much more complex than the real time assignment described above. Our model enables to optimize resources (the vehicles), and to accept only the bookings that the fleet is able to carry out. For instance, several motorcycle-taxi companies in Paris use our tool and our planning solution.</p><p>From an algorithmic perspective, assigning rides in advance means working on planning. For this, we save all booking requests made in advance (up to 20 minutes before the departure time) and then we assign them all at once by building drivers’ schedules.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nncCoTGlmpl4a3kRrNybpQ.png" /></figure><p>Here’s what happens in detail when it comes to planning construction:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YexBFNSdhmSCTKpV21g4bw.png" /></figure><p>This case is relatively simple because here we just have to solve a “static” problem, where we compare all drivers’ shedules to all booking requests to arrange them in the best way.</p><p>Nonetheless, if we want to keep some flexibility (and accept bookings made after the initial construction of the planning), the problem becomes dynamic, and by consequence more complex. In fact, as soon as a new booking is made, we have to recalculate all the plannings, and if necessary, make changes on the plannings by re-assigning the rides.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6IWQwjs2LD5TH7FP7CfcUw.png" /></figure><p><strong>Creation of a hybrid model</strong></p><p>In a nutshell, we find ourselves here facing two opposing ways of dealing with bookings in advance, depending on the service available to make ends meet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*imIaNr7oPM_j99WziR7xPw.png" /></figure><p>Our vision at Yuso is not to consider these two dispatch models as contradictory, but more as complementary, and to use them both in our tool according to the economic relevance of the various booking requests received. For example, this will allow a transporter to assign in advance only the most critical rides, or those that are economically interesting, and to treat in real time rides that are less complicated to assign (for example, bookings that start from train stations or airports).</p><p>To actually implement this hybrid model, we have introduced in our dispatch tool the notion of “clustering”, which will basically consist in forming groups of rides that “match together” and that make sense economically. Accordingly, as soon as a cluster of rides is identified, it will be assigned to a driver and added to his planning.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U1DVruIZ0usmmSxngsHOKw.png" /></figure><p><strong>Towards a more optimized dispatch</strong></p><p>As we have seen throughout this article, these dispatch models are interesting to optimize the transport of people in taxis or PHVs. However, if we want to push optimization even further, and set up a logic of pooling rides (for a delivery or ride-sharing service for example), we have to add a new parameter that complicates even more the equation: the filling rate.</p><p>Considered more important than the approach time and kilometers driven while vacant, the filling rate is essential from an economic point of view, because the absence of mutualization often leads to income losses. Thus, in the case of a pooled ride, it makes no sense to consider the closest vehicle as the most optimal. We have to consider above all the predefined itinerary, the constraints of detours and schedules and obviously the filling rate of the considered vehicle… More criterias to be taken into account means more calculations to be made, and newer challenges in terms of dispatch, but this is a subject that we will discuss in the second part of this article!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=455893821fa" width="1" height="1" alt=""><hr><p><a href="https://medium.com/yuso-articles/the-puzzle-of-dispatch-part-1-advance-booking-455893821fa">The puzzle of dispatch, part 1: advance booking</a> was originally published in <a href="https://medium.com/yuso-articles">Yuso blog articles (in English)</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>