On DICE-free Smart Cities, Particulate Matter, and Feedback-Enabled Access Control

1 On DICE-free Smart Cities, P articulate Matter , and Feedback-Enabled Access Control Panagiota Katsikouli, Pietro Ferraro, David T imoney , Marc Masen, Robert Shorten Abstract —The link between transport related emissions and human health is a major issue f or city municipalities worldwide. PM emissions from exhaust and non-exhaust sources are one of the main worrying contributors to air -pollution. In this paper , we challenge the notion that a ban on internal comb ustion engine vehicles will result in clean and safe air in our cities, since emissions from tyres and other non-exhaust sour ces ar e expected to increase in the near future. T o this end, we present data from the city of Dublin that document that the current amount of tyre- related PM emissions in the city might already be above or close to the levels deemed safe by the W orld Health Organization. As a solution to this problem, we present a feedback-enabled distributed access control mechanism and ride-sharing scheme to limit the number of vehicles in a city and theref or e maintain the amount of transport-related PM to safe levels. I . I N T R O D U C T I O N The link between transport related emissions, and human health, is a major issue for city municipalities worldwide. Diesel and other internal combustion engine based motor vehicles are considered to be the major culprit in this regard as they are associated with the generation of a number of harmful emissions. Apart from the link to global warming through the generation of carbon-dioxide, such vehicles are also known to produce other airborne pollutants such as nitrogen oxides, ozone, benzene, carbon monoxide, and particulate matter (PM) of varying size, all of which are considered harmful to human health. In a recent global revie w [1], it is stated that air pollution, in general, could be damaging ev ery organ and ev ery cell in the human body , showing a potential link between toxic air and skin damages, fertility , asthma and allergies to children and adults. One of the main reasons behind all this is considered to be PM emissions. PM is a generic term used for a type of pollutants that consist of a complex and varied mix of particles suspended in air . Among all the airborne pollutants PM is particularly worrying due to its ability to enter the bloodstream and reach major organs in the human body . There is rich literature documenting the link between PM and its effects on human health [2]–[7]. In particular, the W orld Health Organization reports that “adverse health effects of PM are due to exposure over both short (hours, days) and long (months, years) terms and include respiratory and cardiov ascular morbidity (aggrav ation of asthma, respiratory symptoms, increase in hospital admissions), as well as mortality from cardiov ascular and respiratory diseases and This paper has been submitted to the IEEE Access under the title Dis- tributed ledger enabled control of tyr e induced particulate matter from lung cancer” [8]. Smaller PM particles tend to be more harmful to humans compared to lar ger ones, as they can trav el deeper into the respiratory system [4], [8]. Some of the health effects related to PM include oxidativ e stress, inflammation and early atherosclerosis. Other studies ha ve shown that smaller particles may go into the bloodstream and thus translocate to the liv er , the kidneys or the brain (see [9] and references within). Transport related emissions are a significant contributor to airborne PM lev els that harm our health. In a recent study [10], it is sho wn that living near major roads (i.e. near emissions from vehicles) is associated with increased risk of dementia. The reduction of air quality and population exposure to harmful pollutants as a result of road passenger transportation is discussed in [11], in a case study in the Greater T oronto Area. The survey described in [6] focuses on air pollution originating from non-exhaust emissions such as brake and tyre wear , and highlights the related impact to human health as well as the significance of particulate matter reduction. Roughly speaking, three a venues are being explored world- wide in the fight against urban pollution: (i) outright bans on polluting vehicles and embracing zero tailpipe emission vehicles in certain city zones; (ii) measuring air quality as a means to better informing citizens of zones of higher pollution [12]; and (iii) de veloping smart mobility de vices that seek to minimise the ef fect of polluting de vices on citizens as the y transport goods and indi viduals in our cities [13]–[15]. Option (i) whereby ultra-low emission zones are created by banning internal combustion engine (ICE) based vehicles in certain areas, in addition to embracing electric vehicles (EV’ s), has gained much traction worldwide and is being proposed for adoption in cities such as London and Dublin. Apart from the reduced tailpipe pollutants, an additional attraction of the switch from ICE to EV , is that it is beneficial from the perspectiv e of global warming (reduced carbon dioxide), provided that the energy delivered to the EV’ s can be sourced in a green manner . Thus, reducing our dependency of ICE based vehicles would appear to be very beneficial; not only does the strategy achie ve cleaner air b ut we also potentially tackle climate change through reduced production of carbon dioxide. A major objectiv e of this paper is to challenge the current focus on tailpipe emissions. T o av oid any misunderstanding , we wholeheartedly endorse a reduced dependency on ICE based vehicles. Howe ver , the contemporary narrati ve is based on tailpipe emissions only , and while it is indeed true that EV’ s are zero tailpipe emission vehicles, the tailpipe is only one source of PM. Thus replacing one type of vehicle fleet with another type of vehicle fleet may not result in cities with safe lev els of air quality , especially if the non-tailpipe sources of PM are significant. One contribution of this work is to use data from Dublin to argue that PM lev els from tyres alone may be above that which is deemed safe by the W orld Health Organisation (WHO). While we are by no means the first to argue that tyres are an important source of PM (see in particular the excellent report [16]), we strongly belie ve it is important that stake-holders be reminded of this message, particularly in the context of the current transitioning from ICE-vehicles tow ards an electrification of the vehicle fleet. The second part of our paper describes a distributed access control mechanism that both regulates tyre-based PM generation, and provides fair access to a city zone for a set of competing vehicles. The paper is organised as follo ws. In Section II we discuss tyre-wear related PM emissions, supported by numbers from various sources, with a focus on Dublin in Ireland. W e revie w traf fic produced air-pollution mitigation measures in Section III. Our Access Control Mechanism is presented in detail in Section IV. W e simulate our system and present results in Section V and conclude the paper in Section VI. I I . E L E M E N TA RY C A L C U L A T I O N S Particulate Matter: PM is the product of brake and tyre wear from vehicles as well as a by-product of the engine combustion process. The most common classification of particulate matter is according to size: P M 10 for particles with at most 10 µm diameter , PM 2 . 5 for particles with at most 2 . 5 µm diameter, and ultrafine particles which have a diameter of less than 0.1 micrometres. Smaller PM particles tend to be more harmful compared to larger ones as they are able to get deeper into the respiratory system with ultrafine particles being able to get into the bloodstream and therefore translocate into vital organs such as the liver , the kidneys and the brain. Designated safe levels of PM: According to the WHO, for P M 2 . 5 , the daily maximum deemed safe level on average is 25 µg /m 3 , whereas the annual maximum permitted lev el is on av erage 10 µg /m 3 . For PM 10 , the maximum permitted lev els are on av erage 50 µg /m 3 and 20 µg /m 3 on a daily and annual basis, respectiv ely . In general, non-exhaust emissions (including brake and tyre wear, road surface wear and resuspension of road dust) resulting from road traffic, account for over 90% of P M 10 and over 85% of P M 2 . 5 emissions from traffic [17]. Appr oximate guess of airborne PM in Dublin: T o parse these numbers in terms of tyre abrasion for a city with a high volume of cars, we note that in 2014, nearly 28,000 tonnes of tyre w aste was managed in Ireland [18]. Using publicly available data from the Central Statistics Of fice [19] in Ireland, in 2018 approximately 540,000 pri vate cars were continuously activ e in Dublin throughout the year with an av erage distance travelled of approximately 15,000 km per vehicle. W e assume that approximately 1/3 of the vehicles (i.e. c. 170,000 vehicles) will change their tyres in a year in Dublin 1 . Depending on the type of the tyre and the road conditions, a vehicle (i.e, 4 tyres) loses 50 − 240 mg /km in mass [20], which accounts for 4-6 kg of tyre mass lost before tyres are changed. By considering 4 kg of tyre mass lost per vehicle, we estimate that in Dublin, in 2018, at least 680,000 kg of tyre mass was wasted, 10% of which goes airborne [16], [20] as PM. That corresponds to approximately 68,000 kg of particulate matter in a year, or 185 kg per day , in the city of Dublin. [20] states that approximately 50% of the PM 10 emissions (not specifically to air) fall in the P M 2 . 5 category . [21] reports that c.90% of airborne tyre wear particles are smaller than 1 micrometer in diameter (that is, in the P M 2 . 5 category). In [9], references of pre vious studies state that 3-7% of tyre wear particles contribute to airborne P M 2 . 5 . Rubber in road dust: Finally , to determine the presence of tyre residuals in road dust in an urban environment, and thereby provide an experimental estimate of the relativ e proportion of rubber vs. non-rubber particles, road dust sam- ples were collected from seven locations in central London: Brechin Place, Jay Mews, Exhibition Road, Bayswater Road, by Queensgate Station, Cromwell Road and W estway Road- protect 2 . These locations include major thoroughfares as well as residential streets and an ’average’ urban road dust sample was subsequently created by mixing these v arious collected samples. In order to measure the amount of tyre dust we employed thermogravimetric analysis [22], to measure mass change with temperature increase. The results of this analysis on the aggreg ated road sample are depicted in Figure II. T yre rubber ignites at approximately 700 ◦ C , and at this temperature the thermogravimetric curv e shows a drop of 6 . 6% . This indicates that the mass fraction of rubber (or a similar material) in the collected road dust is approximately 6 . 6% . I I I . A I R P O L L U T I O N M I T I G ATI O N M E A S U R E S Among the early solutions to reduce traf fic related air -pollution has been the application of non-thermal plasma to diesel cars [23]. Similar solutions include the application of catalytic filters [24] for reduced exhaust fumes. Such solutions ho wever , fail to address the non-exhaust emissions from diesel and non- diesel vehicles. The potential of road sweeping and washing to reduce non-e xhaust related emissions was presented in a study in the Netherlands in 2010 [25]. The authors, although they identify non-exhaust emissions as the main source for coarse PM in urban areas, conclude that their approach does not hav e 1 A tyre is changed when it has reached a tread wear of approximately 2 mm (or 1.6 mm as is the legal minimum). This translates into approximately 35,000 km of trav elled distance per v ehicle; howe ver , depending on the dri ving conditions, the travelled distance before a tyre is changed can vary from 10,000 km (harsh braking and acceleration, constant change of gears) to 80,000 km (perfect dri ving conditions and fav ourable road and weather). 2 Data collected at Imperial College London by S. Anderson, M . Mallya, H. Richardson, C. Ching. Fig. 1. Thermogravimetric analysis of road dust aggregated over all sites. a significant reduction in non-exhaust emissions. The benefits of ride-sharing to the en vironment have been discussed in various studies, such as [26]–[29]. Howe ver , these studies do not take a dedicated interest to non-exhaust emissions, but rather , to fuel consumption reduction. Fuel consumption reduction has been addressed with route suggestion solutions in [30]–[32], for trucks and vehicle fleets. In [31], the authors present a linear programming solution to the T ime-Dependent Pollution-Routing Problem. Fleets of vehicles are re-routed depending on traffic, and speeds are recommended based on emissions, driver costs, traffic and peak hour information. As a solution, the authors introduce a departure time and speed optimization algorithm. A similar approach for optimisation of fleet size is proposed in [32]. In the same spirit, authors in [33] study a variety of measures, such as traffic control, ban of heavy duty vehicles (HDV) and speed restriction, in order to achieve reduction of traffic related emissions. Traf fic control (simulated simply by reducing traffic by 20%) and HD V banning ha ve a significant reduction in air-pollutants (20% to 23%), whereas speed control exhibits increase in PM emissions, due to HD V . Last but not least, use of electric vehicles, as an alternativ e to diesel and petrol ones, has been suggested for the reduction of traf fic related air pollutants. In a feasibility study in Canada and Italy [34], the use of electric cars and electric motorcycles shows a reduction in CO 2 emissions, ho we ver , the study ignores non-exhaust related emissions, which are as relev ant to electric vehicles as they are to diesel and petrol ones [17]. Another study in the city of Dublin [35] uses home/work commute and traffic related data to study a number of electric vehicle market penetration scenarios and e valuates the emission decrease under each of them. Howe ver , only tailpipe emissions are taken into consideration, again o verlooking brak e and tyre wear and other non-exhaust emissions. As opposed to the majority of works that address the reduction of road traf fic related emissions, we propose a traf fic control and ride-sharing scheme, that reduces the amount of cars in the streets, and therefore the tyre-related emissions, as well as other non-exhaust and exhaust emissions. I V . F E E D BA C K - EN A B L E D A C C E S S C O N T R OL The pollution mitigation mechanisms discussed in the previous section, and the mov e from ICE to EV’ s that is so popular in many cities globally , is based on the assumption that the principal source of pollution is tailpipe in origin. As we have discussed in the previous section this assumption is at best only partially true and tyres, brakes, as well as road abrasion, may contribute significantly to PM generation. Additionally , EV’ s are generally heavier than ICE vehicles, potentially affecting tyre wear negativ ely . This means that the amount of emitted non-exhaust PM might actually ev en be elev ated for EVs. Therefore, one must look for alternati ve mitigation mechanisms to combat these sources of PM generation. Apart from the obvious mov e from pri vate to public transport or other modes of transport such a cycling and scooters, the only real viable mechanism is to develop an access control mechanism that is based on a feedback control strategy to regulate the safe levels of PM. It is one such strate gy that we no w de velop. Specifically , our objective is to maximise both the number of cars and people entering the city centre each day , while maintaining the tyre-generated PM emission lev els significantly belo w the maximum permitted lev els. The idea is to orchestrate an access control scheme so that it encourages ride-sharing. The access mechanism works in a simple way: at each day passengers are assigned to cars (driv ers) through a matching method. Then, cars who want to have access to the city center are picked randomly using a probabilistic method that ensures fairness and priv acy to each user and which is based on occupancy . See Figure 2, for a visual depiction of this scheme. The rationale behind the choice of a probabilistic method instead of a deterministic one, like a water-filling algorithm, lies in the fact that the latter can be quite inefficient from the single user perspecti ve. In order to use an access control scheme, an agent (driv er or passenger) would typically buy a monthly or yearly access pass. This ticket provides them with the opportunity of competing with other users to access the city , either as a dri ver or as a passenger . Consider no w the example of parents, that have to take their children to school outside the city centre before travelling into the city centre in the morning: e ven though they paid the same amount for a monthly or yearly parking ticket as e veryone else, in a deterministic system they alw ays ha ve a greater chance of missing out the chance of having access to the center of the city , as they arrive later than e veryone else. Using a probabilistic system, as the one described in [36], we are able to guarantee equality in regards to access for all users ov er the long-term period of validity of their pass, irrespectiv e of their constraints. For this Access Control Method, we assume that the controlled region (referred to as R ) can accommodate up to N vehicles per day , decided so that the tyre-related PM emissions are kept at low le vels, ensuring thus that in general PM emission Matchmaking Probabilistic Selection C P M A Fig. 2. Block model of the access control algorithm. First cars and passengers (respectiv ely sets C and P ) get matched by a matchmaking algorithm. Once this process is over , each car and the corresponding passengers (set M ), are randomly picked by a probabilistic method. The resulting set of cars, A represents the cars and passengers who are granted access to the city for the day . lev els will remain low . There are mainly two challenges to make this method work efficiently . Q1 Compliance: How does one make sure that users comply with the matchmaking scheme, after access has been granted? Q2 F air access: How does one ensure that each dri ver is granted access to R fairly with respect to other users (for instance, keeping the amount of av erage access the same among all cars)? W e answer these questions in detail in the following subsec- tions. A. Ride-Sharing Compliance In a Ride-Sharing scheme, as the one described abov e, one of the crucial elements to make the architecture work is to ensure that both driv ers and passengers comply with the matchmaking system. If users are not somehow punished for negati ve behaviour , they might be inclined to cheat the system to maximise their own personal advantage, which in turn might lead to sub-optimal results and to a poor Quality of Service (QoS) ov erall. As an example, in order to increase the probability of gaining access, a dri ver might accept as many dri vers as possible and then refuse to pick them up; on the other hand, a passenger might choose to not show up, effecti vely wasting time and resources (the assigned seat). In this context, on the basis of the work done in [37] we propose the use of a digital token as a bond, or digital deposit, to ensure that passengers and drivers comply with their respectiv e social contract (the matchmaking system). The risk of losing a token is then the mechanism that encourages agents to comply with these social contracts. There are multiple practical ways to implement this system: a possible example could be to ha ve each user equipped with a digital wallet and the only way to participate to the matchmaking system is to have enough tokens to use as a bond. Another way could be to link the tokens to real money , so that losing a certain amount of them would result in a real economic loss for the agent. Note that the pricing of such tokens is beyond the scope of this paper and is dealt with in [37]. The simple idea is that, whenev er a passenger is matched with a driver , they both agree on a specific pick up point and on a time window . Once the passenger gains access to the city center , all the agents inv olved ’ deposit’ a token to the designed pick up point (notice that this process is repeated between each dri ver and passenger, therefore a driv er will deposit an amount of tokens equal to the number of passengers they are carrying). Then, in order to retrieve their token each agent needs to be physically present at the pick up point, in the designed time window . If unable to do so, the agent will forfeit the possession of the token that can be retrie ved by any other passenger/dri ver present at that time and place. T o hav e a better understanding of this process refer to Figure 3. In what follows, we propose the use of a permissioned Distributed Ledger T echnology (DL T) strategy to implement the proposed access control scheme. The acronym DL T is a term that describes blockchain and a suite of related technologies. From a broad perspecti ve, a DL T is nothing more than a ledger held in multiple places, and a mechanism for agreeing on the contents of the ledger , namely the consensus mechanism. While this technology was first discussed in Nakamoto’ s white paper in 2008 [38], the technology has been used primarily as an immutable record keeping tool that enables financial transactions based on peer-to-peer trust [39]. In order to reach consensus, architectures such as blockchain operate a competitive mechanism enabled via mining (Proof- of-W ork), whereas architectures such as the IO T A T angle [40] based on Directed Acyclic Graph (DA G) structures often operate a cooperative consensus technique. The concept of using tokens to mark specific points where conditions are to be met, perfectly conforms with a DL T -based system. In fact, it is natural to use distributed ledger transactions to update the position of the tokens and to link them to the points of interest and associated data, using transactions (this can be done, for example, using smart sensors linked to digital wallets, as shown in Figure 3). On top of that, a DL T -based system brings a number of adv antages as a byproduct of its application to the smart city domain: • Privacy : In DL Ts, transactions are pseudo- anonymous. This is due to the cryptographic nature of the priv ate address 3 , which is less re vealing than other forms of digital payments that are uniquely associated with an individual [41]. This does not mean that DL Ts users’ identities are completely anonymous, especially in architectures in which it is possible to follow the trail of transactions among addresses. At the same time though, DL T systems are pseudo-anonymous in the sense that they manage to hide the details of single users and through randomization of the address they can make it difficult for attackers to trace the transactions. Therefore, from a priv acy perspecti ve, the use of DL T is desirable 3 https://laurencetennant.com/papers/anonymity-iota.pdf A B @ D R R @ @ @ A B @ D R R @ A B @ D R R @ @ @ (a) (b) (c) Fig. 3. (a) Driver D deposits tokens for initiating a contract that they will pick up passengers from pick up points A and B . At the same time, the two passengers deposit their tokens for appearing at the pick-up points. (b) Driver D appears at pick-up point A and collects the passenger from there, therefore, both the driver and the passenger retrieve their tokens for complying with the system. (c) Similarly , tokens are retrieved by the driv er and the second passenger for both appearing at pick-up point B . At each stage the mo ving tokens are represented in green. in a smart mobility scenario. • Ownership : T ransactions in the DL T can be encrypted, thus allowing every issuer to maintain ownership of their own data. In the aforementioned setting, the only information required to remain public is the current ownership of the tokens, whereas auxiliary information (e.g., user quality of service, statistics on the usage of the system) can be encrypted. This information can later be monetized for the benefit for the data o wner . • Micr otransactions : Due to the amount of vehicles in an urban en vironment, and due to the need of linking the information to real time conditions (such as traffic or pollution levels), there is the demand for a fast and large data throughput. Furthermore, the DL T system needs to be designed in a way such that whenev er a user issues a token as a bond, that same user can retriev e the token if and only if they are present at the pick up zone at the designed time. T o do so we make use of the same mechanism and architecture proposed in [42]: namely a Proof of Position (PoP), D AG-based DL T called Spatial P ositioning T oken (SPT oken). Unlike other DL Ts, in which each user has complete freedom on how to update the ledger with transactions, the SPT oken network has a regulatory policy based on the physical positions of agents. This feature allows for a number of different uses: it can be employed to prevent agents to add transactions that do not possess any rele vant data (since transactions can be encrypted) [42] or, as in this specific paper , it can be used to make sure that an agent satisfies certain conditions. Therefore, as a validation mechanism, SPT oken makes use of PoP to authenticate transactions. In other words, for a transaction to be authenticated, it has to carry proof that the agent was indeed at the pick up point, at the designated time. This is achiev ed via special nodes called Observers (see Fig. 3). Each observer is linked to a physical sensor in a city and it acts as a witness for the transaction. A sensor can be a fixed piece of infrastructure, or a trusted vehicle whose position is verified. As soon as a car is granted access to R , each user will deposit their tokens at the designated pick up zone. As soon as an agent reaches in time their pick up point, where one or more of her tokens are av ailable to be picked, a short range connection is established (e.g., via Bluetooth) with the observer (whose job is to authenticate the transaction) and the token is transferred back to the owner’ s account. Refer again to Figure 3 for a better understanding of this process. This mechanism ensures that users have to be physically present in the interested locations to be able to retrie ve their bond. This further authentication step makes SPT oken a permissioned D A G-based DL T (similar to permissioned blockchains [39]), i.e., a distributed ledger where a certain amount of trusted nodes (the observers, in this case) is responsible to maintain the consistency of the ledger (as opposed to a public one, where security is handled by a cooperati ve consensus mechanism [37]). Comment : Before continuing, we want to stress that very often in the context of smart cities , algorithms assume full compliance with policies that are designed to optimise the resource allocation. T o assume that a human agent would not break rules, especially if an individual profit can be made, is a very strong hypothesis that if relaxed might lead the whole system to fail and to produce less than optimal results. Therefore, it is the authors’ opinion that the use of a compliance system is of paramount importance in the setting described so far , if efficienc y is to be achie ved. The issue of compliance is often overlooked. B. Mechanism Description W e consider now the problem of allocating a certain amount of resources (i.e., permitted number of cars) among a set of agents (i.e., driv ers and passengers using the scheme). The proposed method is inspired by the algorithm presented in [36], appropriately adjusted to the requirements of our ride-sharing scheme. W e consider the following scenario. There is a population of size n of citizens participating in the scheme, who request to commute to R on a daily basis. The controlled region can accommodate up to N vehicles per day . W e assume n > N and the population could be either passengers or driv ers. W e assume that there is a fleet of N 0 > N electric vehicles in the scheme that are requested by the population for access in R , with n > N 0 . W ithout loss of generality and to facilitate presentation of our mechanism, the entities driver and car are considered equiv alent and the corresponding terms are thus used interchangeably . As already mentioned in a previous section, our method is organized in two phases : matchmaking and probabilistic access. During matchmaking, we match passengers with dri vers and group them into cars. The matching can happen in a number of ways, depending on the specific requirements of those who apply the system. For example, passengers could be matched with drivers based on proximity of their departing/arriving area, or based on a preference priority ranking that drivers/passengers maintain for each other . In our simulations we take a simple approach and match passengers randomly with drivers (and subsequently with cars), as long as there are av ailable seats in the vehicles, taking into consideration the frequency at which a particular passenger has been assigned a seat in the past. That is, if a passenger has been assigned a seat less than 50% of the time, then they are giv en priority to take a seat in a car , otherwise, they are not given priority . After the matchmaking is complete, each car is assigned an access probability based on its occupancy records. All cars with high enough probability , are permitted access to the city center . W e present the technical details of this procedure, ne xt. In our system, we will use k to denote number of days (i.e., k = 0 , 1 , 2 , 3 , . . . ). For ease of interpretation we assume that access is granted on a daily basis to each user, b ut the algorithm is not affected by this assumption. Then, X i ( k ) is the state v ariable associated with each driv er; it takes the v alue 1 if the i th driver is giv en access to R on the k th day and zero otherwise. Thus, X i ( k ) is the average access for the i th driver up to the k th day , defined as X i ( k ) = 1 k + 1 k X j =0 X i ( j ) . In the above context, let z i ∈ [0 , 1] represent the frequency of accessing the city for a car i , and f i : [0 , 1] → R be a con ve x cost function associated with it, representing the car’ s priority during the second phase of our mechanism. In this context the shape of this function can take into account a variety of factors: the amount of money paid for the pass (e.g., premium and standard account), the amount of public transportation av ailable in the area where this user lives or the type of vehicle driv en. Follo wing [36], we are interested in solving the follo wing shared-resource optimization problem, minimize z 1 ,...,z N 0 ∈ R P N 0 i =1 f i ( z i ) subject to P N 0 i =1 z i = N , (1) z i ≥ 0 , i = 1 , . . . , N 0 . Our aim is then to control the value of the variable X i ( k ) (i.e., the access to R , at each time step) in such a way that the av erage access of user i , X i ( k ) , con verges to the optimal v alue z ∗ i , subject to P n i =1 X i ≈ N (notice that we are not requesting the algorithm to exactly match the required amount of cars, at each time step but we are instead interested in obtaining lim k, ∞ P n i =1 X i ( k ) = N ). In order to do so, the probability that at each time step car i gains access to the city center (i.e., X i ( k ) = 1 ) is ruled by the follo wing equations: p i ( k ) , P ( X i ( k ) = 1) = Γ( k ) X i ( k ) f 0 i ( X i ( k )) n i ( k ) c i , (2) Γ( k + 1) = Γ( k ) + α   N − N 0 X i =1 X i ( k )   , (3) where n i ( k ) is the number of passengers carried in car i at time k , c i is the car’ s maximum capacity and Γ( k ) is a global scaling variable, dependent on the parameter α > 0 , whose dynamics ensures p i ( k ) ∈ [0 , 1] , ∀ i, k . Notice that, equation (2) dif fers from the one proposed in [36] by the factor n i ( k ) /c i : since we are interested in maximising the amount of people getting into R (while maintaining the amount of users having access close to N ), this factor ensures that a fully filled car will hav e higher probability to be granted access than an empty one. As a further element, notice that in a classical setting, the presence of Γ( · ) requires the existence of a centralised entity to compute and broadcast this global variable to all the agents in the network. In a DL T -based system, on the other hand, where informations are stored in a public ledger , the v alue Γ( · ) can be computed independently by each user , therefore the algorithm can be executed in a completely decentralised fashion. A discussion on the con vergence of this algorithm is beyond the scope of this paper and the interested reader can refer to [36] for further details. V . S I M U L A T I O N S A N D R E S U LT S W e now present empirical results to illustrate the ef ficacy of the techniques presented in the previous section. In what follows we based our simulations on the recent report [43]. W e assume a city of the size of Dublin in Ireland, with population 1 , 100 , 000 approximately , of which 50 , 000 are considered dri vers and about 400 , 000 are daily commuting passengers 4 . Consequently , we hav e a fleet of 50 , 000 EV’ s, out of which only 40 , 000 are permitted in the city centre R on a daily basis 5 . All users that are not granted access to the city on a E V , are redirected to use public transportation. In our simulations, we set Γ[0] = 1 , that is, the value that the parameter Γ takes the first day of the scheme’ s operation, and α = 0 . 0001 . W e also consider an application period of 360 days, that is slightly less than a year long. For con venience, on the first day of the operation, we consider that all driv ers are permitted access. The simulation results are presented in Figures 4(a) and (b). Fig. 4(a) shows the number of cars that are granted access ev ery day . Although at the beginning the number of cars in 4 In the report [43] it is stated that between 7-10 am, about 210,000 commuters entered the city center . W e make the assumption that in the length of the day that number can potentially double and therefore consider a population of 400 , 000 commuters 5 In the report [43] it is stated that between 7-10am, about 50,000 cars entered the city center. Therefore, we limit the number of driv ers to that number and the number of permitted cars to slightly less than this figure. (a) (b) Fig. 4. (a) Number of cars with granted access in the length of a year (b) Frequency of granted access per user in the scheme area R are above the maximum permitted number ( 40 , 000 ), due to the ef fect of the access control mechanism this value is quickly reduced, stabilizing around the maximum level, on a verage, for the rest of the application period. Notice that if the maximum number of driv ers in the city center was a hard constraint, it would be sufficient to reduce N to take into account the fluctuations around this value. In Fig. 4(b), we show the frequency of being granted access, per user , on av erage over a period of one year . The small variance indicates that each user is granted fair access to the system. Regarding the commuting (shared) passengers, ev ery passengers is granted access more than 1/3 of the time. Figures 5(a)-(e) depict the number of cars with access, when the number of maximum permitted drivers changes and all other parameters in the system remain the same. The plots depict the steady state values. W e observe that in all cases, the number of cars with granted access conv erges to the maximum value, on average. In terms of fair access, we sho w in Fig. 5(f) boxplots of the frequency at which each driver is granted access to the scheme, in the length of a year, with regards to the maximum number of cars permitted in R . As expected, the frequency increases as the av ailable amount of resources increases. W e highlight that in all cases, the variance is very small, meaning that all drivers in the scheme are ensured fair access (i.e., all drivers are able, on av erage, to access the city center the same number of times). Finally to prove the efficac y of our approach in a more dynamic setting, we allow the number of maximum permitted cars to vary during the year . There are many reasons that make this a realistic scenario: the city municipality might wish to increase the number of permitted vehicles for the holiday seasons, or reduce it during heat waves, for example. W e simulate this setting by changing the number of allowed cars during the year and we present the results in Figure 6. Here, for the first two months of the operating period (i.e., 60 days) we gi ve access to N = 50 k vehicles. For the next forty days N increases linearly and for in the interval (of days) [100 , 180] it is set at N = 80 k vehicles. After that, the number of permitted cars decreases linearly again until it is set to the initial value, N = 50 k , for the rest of the operating period. As we observe in the plot, our system reliably controls the access of vehicles, maintaining the number of permitted cars on average at a stable lev el around the set of maximum values. Regarding pollution le vels caused by the P M 2 . 5 pollutant coming just from the tyre wear of vehicles, we present in Figures 7 the amount of particulate matter , depending on two variables: number of cars permitted in a city (Fig. 7(a)) and the volume of road network in a city (Fig. 7(b)). W ith regards to the volume of a city’ s road network, we wish to estimate, very approximately , the air-space in which the airborne PM is dispersed. For this, we assume that the total mass of PM generated per hour becomes uniformly dispersed throughout a volumetric space which is determined by the street length, an av erage street width of 10m and effecti vely enclosed by an average building height of 4m. Furthermore, we assume rather simplistically that the air in this v olume is continuously replenished with an equiv alent volume of fresh clean air at a rate of one air change per hour , in such a way as to maintain a pollution lev el which remains effecti vely constant with time. For the length of the road network, we can compute the total length of the streets in a predefined area in a city . Note that e ven though 4 m is a somewhat arbitrary number for these simulations, the basic points remain valid irrespectiv e of this assumption; that the amount of tyre generated PM can be regulated using our access control method. T o this end, and based on the above assumptions, Fig. 7 depicts the amount of PM per m 3 as a function of the number of vehicles operating in a city , per possible volume of space (computed as described above). In these figures, we depict in green the lev els deemed safe for human health (i.e., the ones below the maximum permitted le vels) and in red the ones exceeding the annual permitted levels. These plots suggest that, with the present situation in Dublin city (that is, 500,000 cars out of which 170,000 change tyres ev ery year, and a space volume of approximately 450 , 000 , 000 m 3 ), the lev els of tyre-wear related P M 2 . 5 emissions are very high. Howe ver , applying an access control scheme that restricts the number of vehicles to at most 100,000 vehicles per day , can maintain the PM le vels at acceptable levels ev en in small size cities with relati vely small volume of space. V I . C O N C L U S I O N S The contributions of this paper are di vided into two sections. In the first one, a detailed data analysis shows suggests that a simple ban on ICE vehicles does not address the problem (a) Max = 20k cars (b) Max = 35k cars (c) Max = 50k cars (d) Max = 75k cars (e) Max = 90k cars (f) Fig. 5. (a)-(e) Amount of permitted cars, for varied values of maximum allowed vehicles. Steady state values depicted. (f) Frequency of granted access over a year , per dri ver , for different setting of number of allo wed vehicles. Fig. 6. Amount of permitted cars, while changing the value of maximum allowed vehicles gradually from N = 50 k to N = 80 k and vice v ersa. of non-exhaust emissions (PM from tyres, in particular). Although there hav e been previous studies that present such numbers for other cities, we emphasise the point that in Dublin, the PM lev els from tyres alone might be above the lev els that are deemed safe by WHO. This provides us with the rationale to introduce, in the second part, an access control and ride-sharing scheme to limit the amount of cars in cities and therefore maintain the amount of airborne PM within safe lev els for our health. This system is designed in such a way to encourage users to comply with the matchmaking scheme and to guarantee fair access to each car . Finally , to validate the proposed algorithm, we make use of extensiv e simulations to show that each user receives fair access to the city centre and that the PM emissions are kept within safe boundaries. As for future lines of research we will further extend the present work by using more complex models for tyre abrasion and airborne dif fusion to obtain more accurate estimates for non exhaust emissions. R E F E R E N C E S [1] Schraufnagel, Dean E. et al., ”Air Pollution and Noncommunicable Disease”, CHEST , V ol. 155, No. 2 , pp. 409-41, 2019 (a) (b) Fig. 7. Lev els of tyre-wear related P M 2 . 5 emissions per number of vehicles operating (a) and volume of space where the matter is dispersed (b) [2] V alav anidis, A., Fiotakis, K. and Vlachogianni, T ., ”Airborne particulate matter and human health: toxicological assessment and importance of size and composition of particles for oxidative damage and carcinogenic mechanisms. ” Journal of Environmental Science and Health, Part C, V ol. 26, No. 4, pp. 339-362, 2008. [3] Gehring, U., Beelen, R., Eeftens, M., Hoek, G., de Hoogh, K., de Jongste, J., Johan, C., Keuken, M., K oppelman, G., Meliefste, K., Oldenwening, M., Postma, D., van Rossem, L., W ang, M., Smit, H., Brunekreef, B., ”Particulate matter composition and respiratory health: the PIAMA Birth Cohort Study . ” Epidemiology V ol. 26, No. 3, pp. 300- 309. [4] European Environmental Agency , ”Air Quality in Europe e 2014 Report [Internet]. EEA, Copenhagen [cited 3 August 2015]. A vailable at: http: //www .eea.europa.eu/publications/air- quality- in- europe- 2014, 2014. [5] Laden, F ., Schwartz, J., Speizer, F . E. and Dockery , D. W ., ”Reduction in fine particulate air pollution and mortality: extended follow-up of the Harvard Six Cities study . ” American journal of respiratory and critical care medicine, V ol. 173, No. 6, pp. 667-672, 2006. [6] Amato, F ., Cassee, F .R., van der Gon, H.A.D., Gehrig, R., Gustafsson, M., Hafner, W ., Harrison, R.M., Jozwicka, M., Kelly , F .J., Moreno, T . and Prevot, A.S., ”Urban air quality: the challenge of traffic non-exhaust emissions”. Journal of hazardous materials V ol. 275, pp. 31-36, 2014. [7] de Miranda, R. M., de Fatima Andrade, M., Fornaro, A., Astolfo, R., de Andre, P . A. and Saldi va, P ., ”Urban air pollution: a representati ve survey of PM 2.5 mass concentrations in six Brazilian cities. ” Air Quality , Atmosphere and Health, V ol. 5, No. 1, pp. 63-77, 2012. [8] WHO (W orld Health Organization), ”Health ef fects of particulate matter . Policy implications for countries in eastern Europe, Caucasus and central Asia” 2013. [9] Grigoratos, T . and Martini, G., ”Non-exhaust traffic related emissions. Brake and tyre wear PM”, Report no. Report EUR, 26648, 2014. [10] Chen, H., Kwong, J.C., Copes, R., Tu, K., V illeneuve, P .J., V an Donkelaar , A., Hystad, P ., Martin, R.V ., Murray , B.J., Jessiman, B. and W ilton, A.S. , ”Living near major roads and the incidence of dementia, Parkinson’ s disease, and multiple sclerosis: a population-based cohort study . ” The Lancet, pp.718-726, 2017. [11] Hatzopoulou, M., and Eric J. M., ”Linking an activity-based trav el demand model with traf fic emission and dispersion models: Transp. [12] Rai, A. C., Kumar , P ., Pilla, F ., Skouloudis, A. N., Di Sabatino, S., Ratti, C., Y asar, A. and Rickerby , D., ”End-user perspectiv e of low- cost sensors for outdoor air pollution monitoring. ” Science of The T otal En vironment, V ol. 607, pp. 691-705, 2017. [13] Herrmann, A., Liu, M., Pilla, F . and Shorten, R., ”A new take on protecting cyclists in smart cities. ” IEEE Transactions on Intelligent T ransportation Systems, V ol. 19, No. 12, pp. 3992-3999, 2018. [14] Gu, Y ., Liu, M., Naoum-Sawaya, J., Crisostomi, E., Russo, G. and Shorten, R., ”Pedestrian-aware engine management strategies for plug-in hybrid electric vehicles. ” IEEE Transactions on Intelligent Transporta- tion Systems, V ol. 19, No. 1, pp. 92-101, 2017. [15] Sweeney , S., Ord ´ o ˜ nez-Hurtado, R., Pilla, F ., Russo, G., T imoney , D. and Shorten, R., ”A context-aw are e-Bike system to reduce pollution inhalation while cycling. ” IEEE Transactions on Intelligent T ransporta- tion Systems, V ol. 20, No. 2, pp. 704-715, 2018. [16] K ole, P . J., Lohr, A. J., V an Belleghem, F . and Ragas, A., ”W ear and tear of tyres: a stealthy source of microplastics in the environment. ” International journal of en vironmental research and public health, V ol. 14, No. 10, 2017. [17] T immers, V . R. and Achten, P . A., ”Non-exhaust PM emissions from electric vehicles. ” Atmospheric Environment V ol. 134 , pp. 10-17, 2016. [18] En vironmental Protection Agency - W aste T yre Statistics for Ireland, 2014. [19] Central Statistics Office - https://www .cso.ie/en/index.html. [20] Grigoratos, T ., Gustafsson, M., Eriksson, O. and Martini, G., ”Experi- mental inv estigation of tread wear and particle emission from tyres with different treadwear marking. ” Atmospheric En vironment, V ol. 182, pp. 200-212, 2018. [21] Fauser , P ., Tjell, J. C. and Bjerg, P . L., ”Particulate air pollution, with emphasis on traffic generated aerosols. ”, 1999. [22] Coats, A. W . and Redfern, J. P , ”Thermogravimetric analysis. A revie w”, Analyst, V ol. 88, No. 1053, pp. 906-924, 1963. [23] Chae, J. O., ”Non-thermal plasma for diesel exhaust treatment. ” Journal of electrostatics, V ol. 57, No. 3-4, pp. 251-262, 2003. [24] Salvat, O., Marez, P . and Belot, G., ”Passenger car serial application of a particulate filter system on a common rail direct injection diesel engine. ” SAE transactions, pp. 227-239, 2000. [25] Keuk en, M., Hugo D. V . d. G. and Karin V . d. V ., ”Non-exhaust emis- sions of PM and the efficiency of emission reduction by road sweeping and washing in the Netherlands. ” Science of the total en vironment V ol. 408, No. 20, pp. 4591-4599, 2010. [26] Fagnant, D. J. and K ockelman, K. M., ”The travel and environmental implications of shared autonomous vehicles, using agent-based model scenarios. ”T ransportation Research Part C: Emer ging T echnologies, V ol. 40, pp. 1-13, 2014. [27] Y u, B., Ma, Y ., Xue, M., T ang, B., W ang, B., Y an, J. and W ei, Y . M., ”En vironmental benefits from ridesharing: A case of Beijing. ” Applied energy , V ol. 191, pp. 141-152, 2017. [28] Concas, S. and Winters, P . L., ”Impact of carpooling on trip-chaining behavior and emission reductions. ” Transportation Research Record, V ol. 2010, No. 1, pp. 83-91, 2007. [29] Chen, P . Y ., Liu, J. W . and Chen, W . T ., ”A fuel-saving and pollution- reducing dynamic taxi-sharing protocol in V ANETs. ” In IEEE 72nd V ehicular T echnology Conference, pp. 1-5, 2010. [30] Suzuki, Y ., ”A ne w truck-routing approach for reducing fuel consump- tion and pollutants emission. ” Transportation Research Part D: Transport and En vironment V ol. 16, No. 1, pp. 73-77, 2011. [31] Franceschetti, A., Honhon, D., V an W oensel, T ., Bektas ¸, T . and Laporte, G., ”The time-dependent pollution-routing problem”. Transportation Research P art B: Methodological V ol. 56, pp. 265-293, 2013. [32] K oc ¸ , C ¸ ., Bektas ¸, T ., Jabali, O. ans Laporte, G., ”The fleet size and mix pollution-routing problem. ” Transportation Research Part B: Method- ological, V ol. 70, pp. 239-254, 2014. [33] Mahmod, M., V an Arem, B., Pueboobpaphan, R. and De Lange, R., ”Reducing local traffic emissions at urban intersection using ITS coun- termeasures. ” IET Intelligent Transport Systems, V ol. 7, No. 1, pp. 78- 86, 2013. [34] Longo, M., W ahiba Y ., and Dario Z., ”T eam play between renewable energy sources and vehicle fleet to decrease air pollution. ” Sustainability V ol. 8, No. 1, pp. 27, 2016. [35] Brady , J. and O’Mahony , M., ”T rav el to work in Dublin. The potential impacts of electric vehicles on climate change and urban air quality . ” T ransportation Research Part D: Transport and En vironment, V ol. 16 No. 2, pp. 188-193, 2011. [36] W . Griggs, J. Y . Y u, F . W irth, F . H ¨ ausler and R. Shorten, ”On the Design of Campus Parking Systems W ith QoS Guarantees. ” IEEE T ransactions on Intelligent Transportation Systems, vol. 17, no. 5, pp. 1428-1437, May 2016. [37] Ferraro, P ., King, C. And Shorten, B., “Distrib uted Ledger T echnology for Smart Cities, the Sharing Economy , and Social Compliance. ” IEEE Access, V ol. 6, pp. 62728-62746, 2018. [38] Nakamoto, S., “Bitcoin: A peer-to-peer electronic cash system. ” avail- able at https://bitcoin.org/bitcoin.pdf, 2008. [39] Puthal, D., Malik, N., Mohanty , S. P ., Kougianos, E., and Das, G., “Everything Y ou W anted to Kno w About the Blockchain: Its Promise, Components, Processes, and Problems. ” IEEE Consumer Electronics Magazine, V ol. 7, No.4, pp. 6-14, July 2018. [40] Zheng, Z., Xie, S., Dai, H., Chen, X. and W ang, H., “ An overvie w of Blockchain technology: Architecture, consensus, and future trends. ” IEEE International Congress on Big Data, pp. 557-564, 2017. [41] Esposito, C., De Santis, A., T ortora, G., Chang, H. and Choo, K. K. R., “Blockchain: A Panacea for Healthcare Cloud-Based Data Security and Priv acy?” IEEE Cloud Computing, V ol. 5, No. 1, pp. 31-37, 2018. [42] Overko, R., Ordonez-Hurtado, R. H., Zhuk, S., Ferraro, P ., Cullen, A. and Shorten, R., ”Spatial Positioning T oken (SPT oken) for Smart Mobility . ” arXi v preprint [43] Canal Cordon Report, ”Report on trends in mode share of v ehicles and people crossing the Canal Cordon 2006-2017”, 2017. Panagiota (Y ota) Katsikouli recei ved the Diploma and MSc in Computer En- gineering and Informatics from the Poly- technic Uni versity of Patras, Greece, in 2011 and 2013 respecti vely , and the PhD in Informatics from the Uni versity of Edinbur gh, Scotland, in 2017. She is cur- rently a post-doctoral researcher with the Uni versity Colle ge of Dublin. Her research interests include human mobility , smart mobility , analytics for mobile data, distrib uted algorithms for mobility data. Pietro F erraro receiv ed a PhD in con- trol and electrical engineering from the Univ ersity of Pisa, Italy , in 2018. He is currently a Post Doc Fellow with the school of electrical and electronic engineering at University College Dublin (UCD). His research interests include control theory applied to the sharing economy domain. David Timoney receiv ed the Ph.D. de- gree in combustion modeling in diesel engines from UCD where he is currently pursuing the degree in mechanical en- gineering. He was with Ricardo Con- sulting Engineers plc., U.K. He joined UCD as an Assistant Professor in 1981, where he has been inv olving in thermodynamics and energy con version systems. His research acti vities hav e been focused on internal comb ustion engines and on transport-related energy topics. He is a member of the Society of Automotive Engi- neers, a Chartered Engineer and a fellow of the Institution of Engineers of Ireland, and a fellow of the Irish Academy of Engineering. Marc Masen is an Associate Professor in T ribology at Imperial College Lon- don and the educational lead in Me- chanical Engineering Design. He holds a PhD from the Univ ersity of T wente, the Netherlands, where he is also a visiting researcher . He is a past Chair of the Institute of Physics T ribology Group and the conference co- chair of the biannual ICoBT conference-series on BioT ribol- ogy . Marc serves on the editorial boards of the Proceedings of the Institution of Mechanical Engineers Part J: Journal of Engineering T ribology , and the journal Biotribology . His research interests include the friction and wear behaviour of viscoelastic materials. Robert Shorten is Professor of Cyber- physical Systems Design at Imperial College London and Professor of Con- trol Engineering and Decision Science at UCD. He was a co-founder of the Hamil- ton Institute at Maynooth Uni versity , and led the Optimisation and Control team at IBM Research Smart Cities Research Lab in Dublin Ireland. He has been a visiting professor at TU Berlin and a research visitor at Y ale Univ ersity and T echnion. He is the Irish member of the European Union Control Association assembly , a member of the IEEE Control Systems Society T echnical Group on Smart Cities, and a member of the IF A C T echnical Committees for Automotiv e Control, and for Discrete Event and Hybrid Sys- tems. He is a co-author of the recently published books AIMD Dynamics and Distributed Resource Allocation (SIAM 2016) and Electric and Plug-in V ehicle Networks: Optimisation and Control (CRC Press, T aylor and Francis Group, 2017)