13 A Study to Promote Car-Sharing by Adopting a Reputation System in a Multi-Agent Context Emilio Picasso Maria Nadia Postorino and Giuseppe M. L. Sarné Universidad de Buenos Aires, Las Heras 2214, DICEAM, University Mediterranea Buenos Aires, 1127, Argentina 89122 Reggio Calabria, Italy Email: epicasso@uca.edu.ar Email: {npostorino, sarne}@unirc.it Abstract—In recent years the increasing rate of vehicular so on) and some operational costs (e.g., gas, oil, service and traffic due to private mobility caused congestion and environ- so on). The car-sharing system (from hereafter only CS) is a mental impacts in urban contexts all over the world. To face solution able to give the same advantages of a personal car such problems an important contribution might be given by transit systems. However, transit systems are characterized by without the aforementioned disadvantages - mainly costs. a discontinuous spatial and time coverage so that other forms More in detail, CS follows a “Car-As-A-Service” paradigm of mobility, like car-sharing, can be an effective complement [17], i.e. it is a membership based service mainly designed for to it by providing the same flexibility and comfort of private both short time and distance trips, which is usually available on cars. Several studies confirmed that car-sharing is almost as demand (or reservation) to all qualified drivers belonging to a highly appreciated as private cars but having the advantage of lower costs. For such reasons, in recent years the car-sharing community [18]. Three types of CSs are commonly identified, market increased continuously as it has been resulting more namely: (i) Peer-to-Peer (P2P): it takes place among private and more attractive for investors, although its market share users offering their personal cars for money (it was the first remains limited. To encourage the car-sharing philosophy, from type of CS and is receiving a new impulse from information one hand traditional car-sharing companies are trying to reduce technologies); (ii) Business to Consumer (B2C): it is made operational costs to offer lower fares and, from the other hand, several individual owners are starting to share their cars suitably available by business companies with the obvious goal to supported by technology. To promote car-sharing activities, in obtain financial benefits; (iii) Not-For-Profit (NFP): in this case this paper a multi-agent system able to monitor car-sharing users’ local communities or social organizations manage a CS service driving habits is proposed. In particular, agents assist users in with the main aim to incentive a sustainable urban mobility. improving their driving, as well as in building their individual The first CS initiatives were in Zurich (Switzerland) in 1948, reputation measures over time. Such reputation scores can be used to allow the access to car-sharing services and personalized Montpellier (France) in 1970 and Amsterdam (Netherlands) in fares. Experiments on real and simulated data are encouraging 1971, but only after 1980 in Europe and in the USA the first of and show the potentiality of this proposal. positive commercial results were achieved. Nowadays, many Index Terms—Car-Sharing, Multi-Agent System, Reputation CSs (mainly B2C), similar for several aspects, are working all System, Business Model. over the world almost exclusively in highly populated cities with significant congestion and parking problems, although I. I NTRODUCTION there are some examples of CS implementations in medium The growing of urban traffic flows due to the increasing size cities. Recently, CSs are receiving a strong impulse by im- number of private cars, which represent the most part of the provements in information and communication technologies, vehicles moving in urban areas, has worsened the citizens which allow specialized companies to connect potential CS quality of life in terms of local environmental effects and users and owners of private cars desiring to rent their personal traffic congestion [1]–[5]. To face such problems, several vehicles when unused [19]–[21]. measures have been implemented by local authorities to reduce Given its increasing relevance, many studies have been the use of private cars usually based on (i) restrictive rules addressed to explore the main issues characterizing CSs as and/or suitable monetary policies (for instance, by adopting (i) users’ response and habits, for instance in terms of usage road tolls, parking fees, limited traffic zones, interchange areas frequency and effects of information technology [22]–[24] (ii) and so on) [6]–[12] and (ii) promotion of transit in urban environmental benefits, e.g. reduction of vehicle kilometers, areas, which represents the most suitable alternative to private accidents, emissions and fuel consumptions, increase in av- mobility [13]–[15]. In this scenario, the main problem is that erage speed [25], [26] and (iii) cost structure and system private cars are generally more appealing than transit in terms organization, including pricing schemes [27], [28] or the of comfort, privacy and flexibility [16]. Indeed, transit is combined use of car sharing and public transport [29], [30]. characterized by discontinuity both in time and in space (it is Even though in recent years CS has been growing in pop- available at a given time and at fixed stops). On the other hand, ularity among consumers and, consequently, among investors, the ownership of a personal car requires an initial investment its market share is limited with respect to other transport to buy it, some mandatory costs (e.g., insurance, taxes and modalities so that its impact on the overall urban mobility 14 is essentially marginal. Therefore, to increase demand for considered CS scenario and Section III presents the proposed and offer of CS services, they should be more attractive for multi-agent architecture, while Section IV describes in detail users and investors. To this purpose, different policies can the activities carried out by the car-agents. The reputation be implemented and combined to support changes in users’ system is described in Section V and the results of the habits [31] as, for instance, discouraging the use of private simulated experiments are presented in Section VI. Finally, cars in specific urban areas, increasing the number of cars in Section VII some conclusions are presented. available for rent [32], enlarging the urban areas covered by CS services [33], improving the financial appeal for all the CS II. T HE P ROPOSED S CENARIO actors by making CS services more affordable for customers Commonly, the industrial costs (C) of a CS activity are and, at the same time, by increasing economical benefits for represented as: investors. C = CM + CO + CP The implementation of suitable actions addressed to act on the financial aspects of the CS market is then crucial. To where: analyze the context, the cost of a CS service depends on sev- • CM is the Marketing cost derived by advertising, promo- eral factors which can be grouped in Marketing, Organization tional events and whichever activity addressed to promote and Production costs (see below), where the last group also the CS use; depends on customers’ behaviors in using the CS services. • CO is the Organizational cost, which mainly involves Even though the effects of the driving features are not costs for employees, buildings, parking areas and similar completely characterized, a general consensus exists on the features; fact that an “aggressive” driving (e.g. speeding up, hard • CP is the Production cost due to management and use braking and so on) has more than one negative effect [34]. of the fleet. In particular, it has direct impacts on costs for service, and The relevance of each component is strictly related to the affects indirectly CS productivity because vehicles should be CS business modality, i.e. P2P, B2C or NFP. stopped for maintenance. By promoting good driving habits, Generally, the entries having the greater impact on the some CS costs could be reduced in order to (i) offer lower Production costs are those for buying or renting the fleet, fares to users [34] by making CS more appealing with respect the vehicle insurance, taxes and finally the costs for the fleet to private cars and/or (ii) increase the financial profits for CS maintenance and cleaning. Costs due to the adoption of infor- activities by supporting existing companies in their initiatives mation and communication technologies are less significant. as well as by attracting new investors and new actors in the As previously introduced, some of such costs are related to CS business. the habits adopted in using the CS vehicles. Reckless or To support this, a possible approach widely exploited in inappropriate driving habits can lead to accidents or abnormal different contexts involving real and virtual communities is consumptions of both mechanical parts and consumables and represented by the adoption of a reputation system, often this implies higher costs due to the service and lower profits combined with agent technologies [35]. Indeed, intelligent for the stopped time of those vehicles. Furthermore, in the software agents can both monitor CS consumers when they case of P2P-CS, inappropriate driving behaviors can also force are using CS services and compute their individual reputation individual owners to avoid sharing their own cars. score. These scores can be used for different aims as, for An important opportunity for promoting CS activities is to instance, to select users admitted to CS services, which is encourage suitable individual driving modalities. Potentially particularly useful to encourage more individual owners to saved money could be used to reduced maintenance costs share their personal cars, or to determine personalized fares and also for awarding, with personalized lower fares, those awarding the better customers. At the same time, agents consumers having appropriate behaviors. Therefore, as a hoped monitoring CS users’ driving activities can support them in result, such behaviors will permit financial benefits for all the improving their habits. CS actors. In this scenario, the main difficulty is the monitoring of The question is “How is it possible to realize this?” Pro- CS customers. However, progresses made in different fields gresses made in computer science, electronic, control systems, as computer science, electronic, control systems, signal pro- signal processing and communications help us to answer the cessing and communications make it not a very complex question above. Indeed, vehicles can be currently equipped task. The adoption of intelligent software agent technology with at least 120 different types of sensors and their num- can help in monitoring users [36], simulating, controlling and ber is increasing quickly (also including those equipments managing transportation networks at different levels of detail allowing the new form of CS-P2P [19]–[21]). Data gathered by providing intelligent decision-making frameworks [35] as by sensors plugged into cars can also be used to analyze well as managing trust and reputation systems [37]. drivers’ behavior. For instance, some insurance companies In this paper, we investigate on the opportunity to support make available sensor-based equipments to be placed into cars CS customers by adopting a distributed reputation scheme and offer lower insurance fees because of the possibility to within a multiagent system in order to promote CS activities. examine in a semi-automatic way the insured driving behavior In the following, Section II provides an overview of the in case of accident. In this study, we associate each vehicle 15 with an intelligent software agent (hereafter simply agent) that autonomous driving. In such a context, some future scenarios analyzes automatically data collected by the vehicle sensors, foresee that within few decades, mainly for safety reasons, classifies each driver based on his/her behavior and assists only autonomous driving vehicles will be legal [41] and, as a him/her in improving his/her driving. consequence, CS and other forms of mobility, for instance taxi Such a classification will be used to compute the drivers’ services, will be different from those that we know nowadays. reputation, which can be defined as: “what is generally However, the current equipment is already suitable for car- said or believed about a person’s or thing’s character or agent activities to be carried out2 . standing” [38]–[40]. In other words, within a community the To realize the agent tasks described in Section III, a given reputation has the meaning of a collective indirect measure CA analyzes the sensor data collected in real time and exploits of trustworthiness deriving by referrals or ratings provided them to: (i) address the driving habits of the CS users on the by the other community members on the basis of their past basis of its data analysis; (ii) compute an overall score for interactions. evaluating the driving style of the current CS session. The computed reputation scores will be used for both As for the first task, we note that some dashboards already allowing/avoiding the access to the CS service and determining provide information to the driver mainly to optimize the personalized CS fares, e.g. a greater or a lower discount on gasoline usage. In our proposal, we suppose that the CA is the baseline price able to give driver information useful to optimize the use of the vehicle under a more general point of view (e.g., gasoline, III. T HE C AR -S HARING M ULTI - AGENT A RCHITECTURE brakes and so on) or, in other words, to optimize his/her In this Section, we provide a short overview about the driving habits. The second task of the CA is addressed to proposed multi-agent platform, which fits all the types of CS compute a feedback, identified as F , exploited by the Agency (i.e., P2P, B2C and NFP). The components of this platform to update the value of the driver’s reputation measure as are (i) a community of agents, named car-agents (CA), each specified in Section V. one associated with a car, and (ii) their Agency. To compute the value of F 3 , let CAi be the car-agent More in detail, for each driving session carried out on the associated with the vehicle i and let uj be a user exploiting associated car, each car-agent provides to: that CS service. Moreover, let Fi,j S be the feedback computed • analyze some data coming from sensors plugged on board by the agent CAi for the user uj with respect to the service in order to classify the driving habits of the current CS S into the domain [0, 1] ⊂ R, where 0 means the minimum user; appreciation for the ui ’s driving habits and 1 denotes the • support the CS user in improving his/her driving style; maximum one. Based on the information sSi,1 , · · · , sSi,n given • compute a score (i.e., feedback) on the basis of its by the n sensors on the vehicle i, the car-agent CAi calculates monitoring activity, which will be sent to its Agency. the feedback Fi,j S . Different strategies and algorithms can In a complementary way, the Agency provides to: be used for computing such a feedback and, therefore, we describe it as the result of a function F( ) depending on the • collect the feedbacks computed locally by the car-agents parameters sSi,1 , · · · , sSi,n in the form: to update the reputation score1 of each user exploiting the CS service the Agency is managing; S • apply specific policies on the basis of the computed Fi,j = F(sSi,1 , · · · , sSi,n ) (1) drivers’ reputation measures (e.g., the Agency allows or denies the access to the CS service, determines person- Note that Eq. 1 is the kernel of the system and its correct alized fares for the CS service and so on); definition could represent a very complex problem. For a • make available some common services to all the CAs low number of parameters also a simple if-then-else approach associated with it (e.g., the agent white pages). could be applied; differently, for a high number of parameters other computational techniques, among which fuzzy-logic or IV. T HE C AR -AGENT ACTIVITY artificial neural networks, appear more suitable candidates for This section describes the car-agent activities introduced in implementing F . the previous section. Each car-agent accesses all the data coming from the V. T HE R EPUTATION S YSTEM sensors plugged on board of its associated vehicle that are In order to compute the drivers’ reputation scores, we useful for its goals. Note that almost all the new cars are designed a specific reputation system, realized by the Agency, provided with a significant number of sensors and processors which satisfies the following three properties, summarized to analyze sensors data. The number of on board sensors in [42]: (i.e., information sources) is expected to grow significantly, similarly to those on vehicles that currently are equipped to test 2 Similar services are already make available by some fleet management softwares in a centralized way. 1 Note that some corrective actions on the reputation scores could be adopted 3 Note that the function F ( ) could not be the same for all the car-agents in presence of events that cannot be gathered in an automatic way as, for managed by Ag so that the score evaluations can not be uniform along all instance, car body damages, interior cleaning and so on (see Section V). the multiagent system. 16 • each involved entity is time persistent, so that for each expensive CS services. For NFP-CS it is worthless interaction an expectation of future interactions always and, therefore, in this case the value of V is set to exists; 1. • reputation ratings about current interactions are captured – ξj is a system parameter ranging in [0, 1] ⊂ R and spread within the involved community; intended to give the reputation system a uniform • reputation ratings about past interactions are used to guide metric by taking into account those characteristics decisional processes about current interactions. of the CS services not intrinsically considered by More in detail, let CAi be the car-agent associated with the the parameter Ci,jS and mainly due to the adoption vehicle i belonging to a CS company managed by the Agency of different policies by the CS companies. and let uj be the user of the CS service S on i. When the CS • Pj is a penalization coefficient for uj ranging in [0, 1] ⊂ S service ends, the car-agent associated with the shared vehicle R used in the case of behaviors/effects not automatically sends to its Agency the feedback Fi,j S for uj , computed based detectable/verifiable by the car-agents4 . For default PjS is on the information gathered by the vehicle sensors during S. set to 1, i.e. absence of penalization for uj . The Agency will exploit Fi,jS to update the reputation score of On the basis of the reputation score the Agency can adopt uj . different fare policies and even deny the possibility of using a More formally, let Rj ∈ [0, 1] ⊂ R be the reputation of CS service to a customer having a very low reputation score. the user uj . After the car-agent CAi has sent to the Agency the feedback Fi,j S for the service S consumed by uj , then the VI. E XPERIMENTS reputation of uj is updated as follows: In this section we present the results of two experiments ad- dressed to verify the effectiveness of the approach previously Rjnew = (α · σi,j S + (1 − α) · Rjold ) · PjS (2) discussed. The first experiment is addressed to verify if the system where: can compute a reasonable feedback. To this purpose, the • α is a system parameter ranging in [0, 1] ⊂ R and ruling OpenXC repository [43] was exploited. It consists of a the behavior of the reputation system. More in detail, it number of anonymous tuple, i.e. the data do not permit weights the relevance of the parameter σi,j S (see below), their association with drivers, referred to different scenarios which takes into account the feedback Fi,j S , in updating and driving habits. More in detail, each tupla is made as the reputation of uj (i.e., Rj ). In other words, the higher {”name”:”string”, ”value”:integer, ”timestamp”:time}, where its value, the lower the sensitivity of the reputation at the first pair identifies the type of information, the second quick changes and vice versa. pair gives its value and the last pair returns a progressive • σi,j is the contribution to the reputation due to S, which S time. As an example, see the tupla above: also takes into account the feedback Fi,j S . More formally, σi,j is computed as: S { ”name”:”accelerator pedal position”, ”value”:2, S σi,j S = Fi,j S · Vi,j · ξi (3) ”timestamp”:1361454211.483000 } where: To solve the problem given by the anonymity of the – Fi,j S is the feedback computed and sent by CAi to OpenXC data, we generated 100 driver’ profiles belonging to the Agency. five driver styles (respectively named very soft, soft, neutral, – Vi,j S is a parameter belonging to [0, 1] ⊂ R and aggressive and very aggressive). Then for each simulated referred to the monetary cost C(S) of the service driver, 10 driving tracks have been built by suitably assembling S (Eq. 4) computed as: the OpenXC data for a global number of 1000 driving tracks. More in detail, each driving category has 20 simulated   1 S if Ci,j = CM ax drivers. Each category is characterized by a different driving   habit in terms of aggressive driving data (e.g., hard accel- S Vi,j = S (4) eration, hard braking and so on) included into the tracks   Ci,j  otherwise data (see Table I) and randomly assigned to each simulated CM ax driver. Therefore, for a specific simulated driver the associated where CM ax (S) is a system threshold representing driving tracks consist of suitable sequences5 of homogeneous the maximum cost for a CS service after which Vi,j S (e.g., “aggressive” or “not aggressive”) OpenXC tuple match- is considered satured. Therefore, the lower the cost, ing the assigned profile. Finally, given the adopted method the lower the effect on the value of the feedback 4 Note that the evaluation of damages as accidents, damages, interior given by the service S. This is a countermeasure introduced to reduce the weight of positive reputation cleaning, penalties and so on, currently have to be necessarily processed by humans. for marginal CS services and then avoid misleading 5 Each sequence consists of 8 tuple and each driving track is different for behavior addressed to consume such reputation for number of tuple, i.e. time length. 17 Driving Category aggressive/non aggressive ratio drivers’ nature and the involved scenario. When the simulation very sof t 1:0 (i.e., only not aggressive actions) sof t from 4:1 to 2:1 starts, all the drivers receive an initial reputation score of 0.5 neutral 1:1 and for each epoch only 20% of the overall number of drivers aggresive from 1:4 to 1:2 is randomly selected. Clearly, the higher is the percentage of very aggresive 0:1 (i.e., only aggressive actions) drivers correctly identified, the higher is the accuracy of the TABLE I T HE ADOPTED AGGRESSIVE / NON AGGRESSIVE DRIVING ACTIONS RATIO proposed reputation system. Both scenarios provided satisfactory results (depicted in Figure 1). In particular, line A shows that 90% of the drivers nature is correctly recognized after less than 80 epochs. to assemble the driving tracks, a timestamps harmonization Note that initially all the normal drivers are recognized and, procedure needed. However, this procedure did not affect the although it is due to the assigned initial reputation score of 0.5, experimental results in any way. this result does not change along all the simulation. Some tests The computation of the feedback F for a driving track (i.e., carried out by adopting a different initial reputation score (e.g., a driver) is computed on the basis of the ratio between the 0.75) led to a similar result. The proposed reputation system driving time (in seconds) and the overall number of aggressive works well also in presence of drivers’ oscillatory behavior actions in that time. In order to identify an aggressive action and 90% of the aggressive drivers have been recognized (line an Artificial Neural Network (ANN) [44] has been adopted, B) after less than 140 epochs also under these particular a tool able to deal with problems denoted by uncertainty and conditions. frequently used in trasportation research [45], [46] The ANN has been set up after preliminary tests that identified the opti- mal pattern structure and the ANN architecture, topology and learning strategy. More in detail, for the ANN training set we used the data coming from the 30% of the generated driving tracks arranged in patterns. Each pattern contains as input 9 data deriving by three tuple 6 and a unique real value ranging in [0, 1] as output data, where 0/1 means minimum/maximum aggressiveness degree. In particular, each tupla consists of an integer number coding the attribute ”name”, the associated value and the time interval occurring with the previous tupla 7 . Fig. 1. Percentage of drivers correctly identified. Scenarios A and B. The ANN model and learning algorithm we identified as the most profitable solutions, are a three-layer ANN trained by a back-propagation (BP) algorithm [44], having 9, 120 VII. C ONCLUSIONS and 1 nodes for the input, hidden and output layers and CS can play an important role to support public and private hyperbolic and sigmoid activation functions for the neurones mobility and contribute to reduce traffic and environmental of the hidden and output layers. The above described trained problems affecting urban contexts. To this purpose, in this ANN recognized aggressive driver actions on the remaining paper we investigated about the possibility of improving driving tracks with an accuracy of over 79%, which can be convenience and profits for CS users and CS suppliers, re- considered a satisfactory preliminary result given the nature spectively. of the exploited dataset. To address these issues we propose the adoption of a The second experiment is addressed to test the effectiveness reputation system implemented by intelligent software agents of the proposed reputation system. To this purpose, 1000 and tested by performing some experiments based on real and simulated drivers have randomly associated with the driver simulated data. The aim is to identify good driving behaviors typologies presented in Table I. Two different scenarios were to reduce CS fees, and vice versa, thus making the system considered: the first scenario (named A) assumes a uniform more attractive for both CS users and CS suppliers. The first drivers’ behavior along all the simulation, while the second experiment exploited real vehicular sensor data to identify the one (named B) is addressed to test the robustness of the repu- driving users’ habits; the second one, based on simulated data, tation system as regards to oscillatory behaviors by assuming verified the effectiveness of the proposed reputation system for that aggressive drivers try to build a positive reputation on two scenarios. The results of these preliminary experiments cheap CS services for consuming it on expensive CS services encourage future researches for further developments of this (25% of the CS services was assumed to be expensive). In the proposal. simulations, the reputation system parameters α, ξ and P were respectively set to 0.15, 1 and 1, while the feedback F and ACKNOWLEDGMENT the parameter V were randomly generated coherently with the This work has been developed within by the Networks and 6 The target tupla and those referred to the previous and following actions. Complex Systems (NeCS) Laboratory - Department DICEAM 7 Note as the first tupla of each driving track is not considered. - University Mediterranea of Reggio Calabria. 18 R EFERENCES [26] J. Firnkorn and M. Müller, “What will be the environmental effects of new free-floating car-sharing systems? the case of car2go in ulm,” Ecological Economics, vol. 70, no. 8, pp. 1519–1528, 2011. [1] C. A. M. Toledo, “Congestion indicators and congestion impacts: a [27] F. Ciari, M. Balac, and M. Balmer, “Modelling the effect of different study on the relevance of area-wide indicators,” Procedia-Social and pricing schemes on free-floating carsharing travel demand: a test case Behavioral Sciences, vol. 16, pp. 781–791, 2011. for zurich, switzerland,” Transportation, vol. 42, no. 3, pp. 413–433, [2] L. Chen and H. Yang, “Managing congestion and emissions in road 2015. networks with tolls and rebates,” Transportation Research Part B: [28] M. Duncan, “The cost saving potential of carsharing in a us context,” Methodological, vol. 46, no. 8, pp. 933–948, 2012. Transportation, vol. 38, no. 2, pp. 363–382, 2011. [3] E. Cascetta and M. N. Postorino, “Fixed point approaches to the [29] S. A. Shaheen and A. P. Cohen, “Carsharing and personal vehicle estimation of o/d matrices using traffic counts on congested networks,” services: worldwide market developments and emerging trends,” Inter- Transportation science, vol. 35, no. 2, pp. 134–147, 2001. national Journal of Sustainable Transportation, vol. 7, no. 1, pp. 5–34, [4] A. Spickermann, V. Grienitz, and A. Heiko, “Heading towards a 2013. multimodal city of the future?: Multi-stakeholder scenarios for urban [30] J. Schuppan, S. Kettner, A. Delatte, and O. Schwedes, “Urban mul- mobility,” Technological Forecasting and Social Change, vol. 89, pp. timodal travel behaviour: Towards mobility without a private car,” 201–221, 2014. Transportation Research Procedia, vol. 4, pp. 553–556, 2014. [5] T. Fontes, S. Pereira, P. Fernandes, J. Bandeira, and M. Coelho, “How [31] J. Kopp, R. Gerike, and K. W. Axhausen, “Do sharing people behave to combine different microsimulation tools to assess the environmental differently? an empirical evaluation of the distinctive mobility patterns impacts of road traffic? lessons and directions,” Transportation Research of free-floating car-sharing members,” Transportation, vol. 42, no. 3, pp. Part D: Transport and Environment, vol. 34, pp. 293–306, 2015. 449–469, 2015. [6] M. Postorino, G. Musolino, and P. Velonà, “Evaluation of o/d trip ma- [32] M. Nourinejad and M. J. Roorda, “Carsharing operations policies: a trices by traffic counts in transit systems,” in Schedule-Based Dynamic comparison between one-way and two-way systems,” Transportation, Transit Modeling: theory and applications. Springer, 2004, pp. 197– vol. 42, no. 3, pp. 497–518, 2015. 216. [33] K. Uesugi, N. Mukai, and T. Watanabe, “Optimization of vehicle [7] S. Ison and T. Rye, “Implementing road user charging: the lessons learnt assignment for car sharing system,” in International Conference on from hong kong, cambridge and central london,” Transport Reviews, Knowledge-Based and Intelligent Information and Engineering Systems. vol. 25, no. 4, pp. 451–465, 2005. Springer, 2007, pp. 1105–1111. [8] N. Paulley, R. Balcombe, R. Mackett, H. Titheridge, J. Preston, M. Ward- [34] Z. F. Quek and E. Ng, “Driver identification by driving style,” Technical man, J. Shires, and P. White, “The demand for public transport: The report, technical report in CS 229 Project, Stanford university, Tech. effects of fares, quality of service, income and car ownership,” Transport Rep., 2013. Policy, vol. 13, no. 4, pp. 295–306, 2006. [35] M. N. Postorino and G. M. L. Sarné, “Agents meet traffic simulation, [9] S. Liu, K. P. Triantis, and S. Sarangi, “A framework for evaluating control and management: A review of selected recent contributions,” the dynamic impacts of a congestion pricing policy for a transportation in Proceedings of the 17th Workshop “from Objects to Agents”, WOA socioeconomic system,” Transportation Research Part A: Policy and 2016, ser. CEUR Workshop Proceedings, vol. 1664. CEUR-WS.org, Practice, vol. 44, no. 8, pp. 596–608, 2010. 2016. [10] M. N. Postorino, “A comparative analysis of different specifications of [36] D. Rosaci and G. M. L. Sarné, “Multi-agent technology and ontologies modal choice models in an urban area,” European journal of operational to support personalization in b2c e-commerce,” Electronic Commerce research, vol. 71, no. 2, pp. 288–302, 1993. Research and Applications, vol. 13, no. 1, pp. 13–23, 2014. [11] A. de Palma and R. Lindsey, “Traffic congestion pricing methodologies [37] J. Granatyr, V. Botelho, O. R. Lessing, E. E. Scalabrin, J.-P. Barthès, and technologies,” Transportation Research Part C: Emerging Technolo- and F. Enembreck, “Trust and reputation models for multiagent systems,” gies, vol. 19, no. 6, pp. 1377–1399, 2011. ACM Computing Surveys (CSUR), vol. 48, no. 2, p. 27, 2015. [12] M. Gibson and M. Carnovale, “The effects of road pricing on driver [38] A. Jøsang, R. Ismail, and C. Boyd, “A survey of trust and reputation behavior and air pollution,” Journal of Urban Economics, vol. 89, pp. systems for online service provision,” Decision support systems, vol. 43, 62–73, 2015. no. 2, pp. 618–644, 2007. [13] J. D. Harford, “Congestion, pollution, and benefit-to-cost ratios of us [39] D. Rosaci, G. M. L. Sarné, and S. Garruzzo, “TRR: An integrated public transit systems,” Transportation Research Part D: Transport and reliability-reputation model for agent societies,” in Proceedings of the Environment, vol. 11, no. 1, pp. 45–58, 2006. 12th Workshop from “Objects to Agents”, WOA 2014, ser. CEUR Workshop Proceedings, vol. 741. CEUR-WS.org, 2011. [14] M. Postorino and V. Fedele, “The analytic hierarchy process to evaluate [40] M. N. Postorino and G. M. L. Sarné, “An agent-based sensor grid the quality of service in transit systems,” WIT Transactions on The Built to monitor urban traffic,” in Proceedings of the 15th Workshop from Environment, vol. 89, 2006. “Objects to Agents”, WOA 2014, ser. CEUR Workshop Proceedings, [15] C. Winston and V. Maheshri, “On the social desirability of urban rail vol. 1260. CEUR-WS.org, 2014. transit systems,” Journal of urban economics, vol. 62, no. 2, pp. 362– [41] D. Newman, “D. Newman, ”Autonomous 382, 2007. Cars: The Future Of Mobility”, available at [16] B. Caulfield, “An examination of the factors that impact upon multiple http://www.forbes.com/sites/danielnewman/2016/09/27/autonomous- vehicle ownership: The case of dublin, ireland,” Transport Policy, cars-the-future-of-mobility/#6d7ca743514a,” 2017. vol. 19, no. 1, pp. 132–138, 2012. [42] P. Resnick, K. Kuwabara, R. Zeckhauser, and E. Friedman, “Reputation [17] J. Agyeman, D. McLaren, and A. Schaefer-Borrego, “Sharing cities,” systems,” Communications of the ACM, vol. 43, no. 12, pp. 45–48, 2000. Friends of the Earth Briefing, pp. 1–32, 2013. [43] Openxcplatform. (2017) http://openxcplatform.com/resources/traces.html. [18] C. Sinclair, “Codes of ethics and standards of practice.” 1993, available [44] S. Haykin, Neural Networks - A Comprensive Foundation. New York: at www.carsharing.org. Macmillan College Publishing Company, 2000. [19] Car2Share, “https://www.car2share.com,” 2017. [45] M. N. Postorino and G. M. L. Sarné, “A neural network hybrid [20] Drife, “http://dryfe.it,” 2017. recommender system,” in Proceedings of the 2011 conference on neural [21] Getaround, “https://www.getaround.com/san-francisco,” 2017. Nets WIRN’10, 2011, pp. 180–187. [22] P. Baltusis, “On board vehicle diagnostics,” SAE, Tech. Rep., 2004. [46] M. N. Postorino and M. Versaci, “A neuro-fuzzy approach to simulate [23] K. M. N. Habib, C. Morency, M. T. Islam, and V. Grasset, “Modelling the user mode choice behaviour in a travel decision framework,” users behaviour of a carsharing program: Application of a joint hazard International Journal of Modelling and Simulation, vol. 28, no. 1, pp. and zero inflated dynamic ordered probability model,” Transportation 64–71, 2008. research part A: policy and practice, vol. 46, no. 2, pp. 241–254, 2012. [24] Telematics, “http://www..com/board-diagnostics-future-vehicle- analysis,” 2017. [25] N. Fellows and D. Pitfield, “An economic and operational evaluation of urban car-sharing,” Transportation Research Part D: Transport and Environment, vol. 5, no. 1, pp. 1–10, 2000.