Current paradigms in intelligent transportation systems

Published in IET Intelligent Transport Systems Received on 11th November 2009 Revised on 12th May 2010 doi: 10.1049/iet-its.2009.0102 ISSN 1751-956X Current paradigms in intelligent transportation systems S.L. Toral 1 M.R. Martı ´nez Torres 2 F.J. Barrero 1 M.R. Arahal 1 1 E.S. Ingenieros, University of Seville, Avda. Camino de los Descubrimientos s/n, Seville 41092, Spain 2 E.U.E. Empresariales, University of Seville, Avda. San Francisco Javier s/n, Seville 41018, Spain E-mail: [email protected] Abstract: Intelligent transportation systems (ITS) constitute today a multidisciplinary field of study involving a large number of different research areas. As a consequence, it is difficult to achieve a structured view of ITS, which is necessary to unify efforts and as guidance for future developments. This study aims to identify the main paradigms in the field of ITS by semantically analysing studies related to this general topic. An understanding about which research is considered valuable by the research community to build upon may provide valuable insights in this field. As a result of the statistical treatment of data, up to 13 paradigms are obtained. The scope of these paradigms and the relationships between them have also been detailed, providing a structured vision of ITS synthesised in a map form. 1 Introduction Intelligent transportation systems (ITS) have been investigated for many years in Europe, North America and Japan, with the aim of improving the safety and efficiency of road transport and environmental conservation. To this end, new technologies and computer power have been applied to freeway, traffic and transit systems [1, 2]. ITS can be considered a global phenomenon, attracting worldwide interest from transportation professionals, the automotive industry and political decision makers [3]. ITS involves a large number of research areas spread over many different technological sectors such as electronics, control, communications, sensing, robotics, signal processing and information systems [4, 5]. This multidisciplinary nature increases the problem’s complexity because it requires knowledge transfer and cooperation among different research areas [6]. One of the main problems of being a complex and multidisciplinary field is the difficulty of dealing with ITS as a development and research area [7, 8]. Traditional topics of interest are changing and new ones are emerging due to the continuous advance of emergent technologies and the economical, social and environmental implications of ITS. The complexity of the topic suggests that it will be of benefit to define a global, structured view. This view will be useful to support close integration of ITS with conventional transportation initiatives as well as to provide guidance for future ITS deployments. Sharing a common structured view will also help the promotion of ITS standards development, the identification and confirmation of needs, problems, objectives and issues, and the alignment of researchers, companies and users for synergy [3]. The idea of obtaining this structured view of a particular research and development area is not new, but nowadays is taking on major importance. The most recent and ambitious attempt has been made by the European commission by launching the technological platform called ARTEMIS [9]. The aim of ARTEMIS is to develop and drive a joint European vision on embedded systems connecting research and development with innovation to align fragmented R&D efforts along common strategic agenda and looking for improvements in European companies’ efficiency and competitiveness. ARTEMIS is following a bottom-up scheme in which the most prominent European companies are getting involved in the definition of this unique European vision through the development of a strategic research agenda (SRA). In IET Intell. Transp. Syst., 2010, Vol. 4, Iss. 3, pp. 201–211 201 doi: 10.1049/iet-its.2009.0102 &The Institution of Engineering and Technology 2010 www.ietdl.org particular, ITS can be located in one of the four applications contexts identified by ARTEMIS SRA devoted to public infrastructure [9]. Although some structured view of ITS has been proposed based on market areas [3] or on databases and cumulative experience [10], this study proposes a quantitative and systematic methodology to identify the main paradigms within the broader field of ITS. The starting point is the information provided by the Institute for Scientific Information’s (ISI) massive datasets. The abstracts of the published papers included in datasets will have been analysed using text categorisations tools to obtain the final paradigms. To meet these objectives, the paper has been structured as follows. In Section 2, ITS will be analysed as a field study, describing previous attempts of structuring this research topic. In Section 3, the proposed methodology will be described in detail, including a discussion about some other approaches. The obtained paradigms in ITS will be presented in Section 4, through the application of the proposed methodology. Finally, the main conclusions of the work are included in Section 5. 2 Analysis of ITS as a field study There are different ways to classify and segment the ITS field. Six major categories were reviewed in [6] from a technological perspective: †Advanced traffic management systems (ATMS), used to improve traffic service quality and to reduce traffic delays. †Advanced travellers information systems (ATIS), used to supply real-time traffic information to travellers. †Commercial vehicles operation (CVO), systems that use different ITS technologies to increase the safety and efficiency of commercial vehicles and fleets. †Advanced public transportations systems (APTS), which make use of electronic technologies to improve the operation and efficiency of high-occupation transports, such as buses and trains. †Advanced vehicles control systems (AVCS), which joint sensors, computers and control systems in driving assistance solutions. †Advanced rural transports systems (ARTS), used to solve problems arising in rural zones (steep grades, blind corners, curves, scarce navigational signs, mix of users, lack of alternative routes). Although this classification is shaped by emerging technologies and is useful from the viewpoint of system designers, some interesting topics related to ITS are excluded. For instance, transport policy and planning, traffic modelling and forecasting, and the sociological and behavioural influences of ITS are not included in the previous classification. The requirements and preferences of ITS users may also be given inadequate attention. Another way of looking at ITS is to consider market areas as a representation of ITS users and operators with similar needs. Nine major market areas, detailed in Table 1,are defined in [3]. Similar classifications can also be found in [10], although they consider up to 13 areas by sub-dividing several of the nine market areas defined in Table 1. As a difference to the previous classifications, mainly focused on one aspect of ITS, an integral analysis of the ITS field is proposed in this paper. The starting point will be the abstracts and keywords of published papers related to ITS included in the ISI Web of Science database [11, 12]. ISI Web of Science includes journals of almost Table 1 Market areas in ITS Market area Goal area 1 traffic management manage the entire road network on behalf of the general public area 2 emergency management respond to incidences and emergencies (fire, police, ambulance) area 3 transportation planning match transportation supply with demand both now and in the future area 4 traveller information supply information to traveller and subscribers area 5 commercial vehicles provide travel information and fleet management services area 6 transit management plan and operate transit systems in both urban and rural areas area 7 intelligent vehicles enhance the capabilities of road vehicles through the use of electronics, sensors, communications and control actuator technologies area 8 incident management concerned with efficiency and safety of the roadway network area 9 payment systems encompass all the people that take money in return for providing a service (toll road operators, transit agencies, car parking operators, etc.) 202 IET Intell. Transp. Syst., 2010, Vol. 4, Iss. 3, pp. 201–211 &The Institution of Engineering and Technology 2010 doi: 10.1049/iet-its.2009.0102 www.ietdl.org each scientific field. Consequently, not only technological and market approaches to ITS will be gathered, but any scientific discipline strongly or weakly related to ITS, such as planning and logistic, psychology or social sciences. The selected papers will be processed using a statistical text categorisation tool and, as a result, major paradigms in the field of ITS will be identified. 3 Methodology The starting point of the proposed methodology consists of paper extraction from ISI databases. Particularly, papers related to the topic ‘intelligent transportation systems’. Instead of analysing the full text, which would be an enormous task, a representative piece of text summarising the whole paper has been selected, i.e. the abstracts and keywords or index terms. An abstract is a condensed version of a paper that highlights the major points covered and concisely describes the content and scope of the writing, while keywords review the writing’s contents in abbreviated form. Keywords are included because they emphasise the content of the paper. Although keywords themselves can constitute an adequate description of the paper content [13], they can also restrict the number of different topics to be obtained. Notice that sometimes keywords must be chosen among a closed list provided by the journal publisher. Consequently, the proposed methodology will consider both abstracts and keywords with the aim of leaving open the number of topics to be obtained. It is implicitly assumed that authors write good abstracts (representative of the content of the paper) and that keywords are carefully chosen. In general, bibliometric research is devoted to quantitative studies of literature. Several empirical methods can be found in the literature. Co-citation methods are perhaps the most employed [14], and they have been frequently used to analyse the intellectual structure of many disciplines [15]. The basis of co-citation methods consists of counting the number of times certain markers occur or co-occur, giving rise to information on such author co-citation [16], journal co-citation, keyword co-citation, etc. [13]. The main drawback of traditional bibliometric techniques, such as author or journal co-citation methods, is that they are not concerned about the content of considered papers but on references usually delayed between 2 and 5 years after a paper is first drafted. Although they lead to interesting results, they do not provide an immediate picture of the actual content of the research topic dealt with in the literature. As a difference, semantic analysis based on co-words analysis (co-occurrences of words in the publications on a given subject) has the potential of solving this kind of problem [13–17]. Semantic analysis usually employs a vector space model [18], in which documents are summarised and represented by vectors of words (term vectors). However, a central problem in this kind of statistical analysis is the high dimensionality of the feature space (one dimension for each unique word). Therefore, it is desirable to first project the documents into a lower-dimensional subspace in which the semantic structure of the document space becomes clear [19]. In the low-dimensional semantic space, the traditional clustering algorithms can then be applied. To this end, spectral clustering [20, 21], clustering using latent semantic indexing (LSI) [22] and clustering based on nonnegative matrix factorisation [23, 24] are the most wellknown techniques. Particularly, LSI decomposes a term document matrix using a technique called singular value decomposition to construct new features as combinations of the original features, significantly reducing the highdimensionality problem of the feature space [25]. Moreover, LSI considers documents that have many words in common to be ‘semantically close’, and ones with few words in common to be ‘semantically distant’. The LSI approach makes three basic claims: that semantic information can be derived from a word-document co-occurrence matrix; that dimensionality reduction is an essential part of this derivation; and that words and documents can be represented as points in a Euclidean metric space. A different approach has been applied in this paper. This approach is consistent with the first two of these claims, but it differs in the third, describing a class of statistical models in which the semantic properties of words and documents are expressed in terms of probabilistic topics [26]. The topic model is a statistical language model that relates words and documents through topics. It is based upon the idea that documents are mixtures of topics, where a topic is a probability distribution over words [26–28]. In this paper, the methodology proposed for ITS paradigms identification consists of a four-step procedure based on the latent Dirichlet allocation (LDA) method of [26] (see Appendix for more details) and illustrated in Fig. 1. Figure 1 Dataflow of the proposed methodology IET Intell. Transp. Syst., 2010, Vol. 4, Iss. 3, pp. 201–211 203 doi: 10.1049/iet-its.2009.0102 &The Institution of Engineering and Technology 2010 www.ietdl.org 1. Information extraction: the extraction process involves access to ISI databases to annotate the abstract and keywords of a paper dealing with the topic ‘intelligent transportation systems’. 2. Ontology: an ontology defines the basic terms and relations comprising the vocabulary of a topic area as well as the rules for combining terms and relations between terms [29].An ontological model of the domain is used as a facilitator throughout all the processes. This step provides a common vocabulary and specifies the semantics of key relationships within the domain. The selection is based on the obtained frequency-of-occurrence-rates of words over the total corpus of extracted texts. 3. Structuring and processing information: in statistical natural language processing, one common way of modelling the contributions of different topics to a document is to treat each topic as a probability distribution over words, viewing a document as a probabilistic mixture of these topics [27]. 4. Categorisation: The algorithm outlined above can be used to find the topics that account for the words used in a set of documents. In this study, every abstract and its associated keywords are considered a document. 4 Paradigms in ITS A total of 1147 papers related to the topic ‘intelligent transportation systems’ has been obtained from ISI databases, covering subject areas like engineering, transportation, computer science, telecommunications, automation and control systems, and operations research and management science. The majority of them belong to the general category of Science & Technology, but a small percentage is associated to Social Sciences and Arts & Humanities. According to the explained procedure, abstracts and keywords have been used as documents in the terminology of semantic analysis. Up to 64 132 words have been extracted from these documents, leading to a vocabulary (non-repeated words) of 5485 words. Working with such an amount of words would be prohibitive in terms of computing time, so ontology is selected to facilitate the implementation of the procedure. The criterion for the ontology selection was based on the frequency of terms in the document collection. Particularly, words that occurred in more than 15 documents were considered [30].Asa result, an ontology of 267 words were obtained. This is the most manual-step of the procedure, because the final result must be supervised to remove words that are not directly related to ITS issues. Table 2 shows the key dimensions used in the topic model. T and ITER represent topic model run parameters. The number of Gibbs sampler iterations was chosen to be ITER ¼200. This is a large enough value to guarantee the convergence of the algorithm [21]. The number of topics was selected using the perplexity value. Perplexity is a standard measure of performance for statistical models of natural language [31, 32] defined by (1). pplex =exp −1 W W n=1 log P(wn|dn)  (1) The role of perplexity has mostly been discussed on an intuitive level as average uncertainty when predicting the next word given its history. Perplexity indicates the uncertainty in predicting a single word. A lower perplexity score indicates better generalisation performance. Perplexity varies from 1 to W; lower perplexity is better, and the maximum perplexity of Wis reached when all words in the vocabulary are equally likely. In our case study, the LDA algorithm was run for a number of topics varying between 1 and 30. The results are illustrated in Fig. 2. The minimum perplexity value is reached for a number of topics equal to 13. Consequently, 13 was chosen as the number of selected topics. Table 3 shows the 13 topics obtained, with the most likely words in each topic, and their probabilities P(w|t). Table 2 Dimensions of the topic model Parameter Description Value Dnumber of documents in corpus 449 Ntotal number of words in corpus 34 132 Laverage length of document in words (L¼N/D) 76 Vnumber of words in vocabulary 5485 Wontology 267 Tnumber of topics – ITER number of iterations 200 Figure 2 Perplexity as a function of the number of topics 204 IET Intell. Transp. Syst., 2010, Vol. 4, Iss. 3, pp. 201–211 &The Institution of Engineering and Technology 2010 doi: 10.1049/iet-its.2009.0102 www.ietdl.org Table 3 Obtained topics and probabilities from the LDA algorithm TOPIC 1 TOPIC 2 TOPIC 3 TOPIC 4 method 0.165 system 0.101 vehicle 0.265 incident 0.079 effect 0.087 service 0.074 vehicles 0.118 performance 0.067 benefit 0.073 systems 0.073 system 0.069 detection 0.065 analysis 0.072 transport 0.059 information 0.069 technique 0.061 methods 0.055 decision 0.054 transit 0.048 traffic 0.053 benefits 0.049 information 0.049 mobile 0.038 freeway 0.049 evaluation 0.047 framework 0.039 reserved 0.027 network 0.047 effective 0.047 support 0.039 automatic 0.025 management 0.047 result 0.037 environment 0.037 vision 0.025 conditions 0.047 methodology 0.032 assess 0.036 infrastructure 0.025 neural 0.040 intelligent 0.031 intelligent 0.032 approach 0.022 system 0.037 station 0.028 services 0.031 rights 0.021 real-time 0.033 potential 0.026 evaluate 0.030 function 0.021 techniques 0.033 effects 0.024 operations 0.023 changes 0.020 characteristics 0.031 accuracy 0.022 impacts 0.020 corridor 0.020 delay 0.031 alternative 0.022 technologies 0.019 motion 0.018 detector 0.027 effectiveness 0.021 analysis 0.018 moving 0.017 presented 0.027 experimental 0.021 weather 0.018 scenarios 0.017 strategies 0.023 TOPIC 5 TOPIC 6 TOPIC 7 TOPIC 8 system 0.191 algorithm 0.187 transport 0.127 traffic 0.294 systems 0.134 algorithms 0.073 transportation 0.126 simulation 0.118 design 0.078 process 0.059 develop 0.089 signal 0.072 communication 0.060 present 0.057 system 0.066 results 0.047 present 0.051 develop 0.049 state 0.053 estimate 0.043 technology 0.046 approach 0.045 development 0.048 result 0.038 position 0.034 application 0.045 planning 0.041 parameters 0.034 highway 0.032 developed 0.044 region 0.039 estimates 0.032 message 0.028 sensor 0.043 architecture 0.038 intelligent 0.030 research 0.024 performance 0.041 intelligent 0.032 arterial 0.028 integrated 0.023 efficient 0.036 deployment 0.031 temporal 0.024 presented 0.020 complex 0.036 developing 0.028 roadway 0.023 implementation 0.020 applications 0.030 program 0.026 intersection 0.022 potential 0.019 transportation 0.026 issues 0.025 microscopic 0.020 communications 0.019 intelligent 0.022 management 0.023 spatial 0.019 global 0.017 scheme 0.022 process 0.022 empirical 0.017 Continued IET Intell. Transp. Syst., 2010, Vol. 4, Iss. 3, pp. 201–211 205 doi: 10.1049/iet-its.2009.0102 &The Institution of Engineering and Technology 2010 www.ietdl.org Table 3 Continued TOPIC 5 TOPIC 6 TOPIC 7 TOPIC 8 context 0.017 processing 0.021 public 0.019 emissions 0.017 wireless 0.016 sensors 0.019 regional 0.019 evaluated 0.014 TOPIC 9 TOPIC 10 TOPIC 11 TOPIC 12 transportation 0.144 control 0.138 network 0.164 model 0.356 transport 0.142 driver 0.092 problem 0.116 models 0.121 system 0.087 system 0.087 dynamic 0.089 traffic 0.067 application 0.070 vehicle 0.067 networks 0.069 develop 0.056 systems 0.068 systems 0.065 problems 0.042 transportation 0.054 location 0.045 result 0.053 present 0.040 approach 0.039 intelligent 0.044 driving 0.052 computation 0.035 modelling 0.036 technologies 0.043 drivers 0.041 solution 0.033 developed 0.035 user 0.042 results 0.039 routing 0.033 transport 0.030 applications 0.040 safety 0.033 guidance 0.030 urban 0.029 advanced 0.025 conditions 0.022 assignment 0.030 intelligent 0.028 research 0.024 significant 0.021 structure 0.028 forecasting 0.025 experience 0.022 behaviour 0.020 optimal 0.027 research 0.024 function 0.020 tracking 0.019 presented 0.024 demand 0.017 requirements 0.019 camera 0.016 real-time 0.022 systems 0.016 quality 0.017 autonomous 0.016 optimisation 0.022 regression 0.015 commercial 0.016 dynamics 0.015 shortest 0.020 present 0.012 solution 0.015 human 0.015 computational 0.020 feasibility 0.012 TOPIC 13 travel 0.205 – – – – – – predict 0.088 – – – – – – information 0.071 – – – – – – times 0.068 – – – – – – prediction 0.045 – – – – – – atis 0.045 – – – – – – result 0.042 – – – – – – traveller 0.040 – – – – – – choice 0.035 – – – – – – results 0.031 – – – – – – estimate 0.025 – – – – – – travel-time 0.021 – – – – – – future 0.020 – – – – – – statistical 0.019 – – – – – – Continued 206 IET Intell. Transp. Syst., 2010, Vol. 4, Iss. 3, pp. 201–211 &The Institution of Engineering and Technology 2010 doi: 10.1049/iet-its.2009.0102 www.ietdl.org Each topic can be derived from its corresponding bag of words, leading to the following topic categorisation list: †Topic 1. Evaluation of effectiveness and benefits of ITS: This topic deals with the analysis of implications and potential use of ITS including new challenges and opportunities, and the anticipation of user behaviour in the process of designing new ITS technologies. †Topic 2. Study of systems and tools supporting decision making and transportation planning: It includes the development of models for travel demand decision and transportation planning modelling tools, technology integration and databases supporting archived data user services. †Topic 3. Automatic vehicle detection systems: Vehicle detection appears to be one of the most promising areas in traffic surveillance and control. This concept entails the detection of vehicles and extraction of traffic parameters in real-time from images generated by video cameras overlooking a traffic scene. †Topic 4. Incident management: Incident management includes emergency response deployment and rerouting to bypass the affected area, impact of freeway lane closures, and incident detection algorithms. †Topic 5. Mobile and wireless communication in ITS: This topic covers intervehicle communication networks, in-vehicle communications ), road-to-vehicle communications, wireless protocols, GPRS and thirdgeneration systems. †Topic 6. Processing algorithms for ITS applications: This topic is devoted to advanced processing algorithms for solving ITS problems, such as path problems in dynamic networks, origin–destination estimation and prediction, and image and video processing algorithms. †Topic 7. Advanced traffic management systems (ATMS): They are focused on the development of ITS to improve safety and quality of service as well as the efficiency of existing roadway utilisation. †Topic 8. Traffic simulation: A traffic simulation system consists of a traffic-flow simulation code, which is able to simulate traffic on a freeway network. It usually considers a microscopic representation (where each individual vehicle is represented) or a macroscopic model capturing traffic dynamics. The purpose of these systems consists of performing traffic-flow simulation for applications like traffic conditions prediction in real-time, traffic control and drivers’ guidance, link travel time calculation and signal control strategy. This topic also covers issues like modelling driver behaviour under the influence of external factors. †Topic 9. Commercial Vehicles Operation (CVO): This topic is focused on the impact of ITS on commercial vehicles and fleets for improving transportation safety and efficiency. †Topic 10. Advanced vehicles control systems (AVCS): AVCS are based on systems that provide increased safety and/or control to the driver either by means of improving the information about the driving environment or by actively aiding the driver in the driving task. They include onboard autonomous intelligent cruise control systems, ABS and traction control systems, active suspension systems, vehicle stability systems, in-vehicle collision warning systems, etc. †Topic 11. Dynamic route selection algorithms: Dynamic route selection problems are search problems for finding an optimal route from a starting to a destination point on a road map within a time limit. Since the time to traverse a link will depend upon traffic volume encountered on that link, link times are dynamic. †Topic 12. Models for traffic demand forecasting: Traffic congestion is a major operational problem on ITS. Reducing congestion effects requires developing models that can accurately predict traffic demand. †Topic 13. Advanced travellers information systems (ATIS): The function of ATIS is to assist travellers with planning, perception, analysis and decision making to improve the convenience and efficiency of travel. The obtained results depict an exhaustive draw of ITS research areas. Some market-specific topics are suppressed compared to previous classification methods described in Section 2, like APTS or ARTS in [6] or emergency management and payment systems in [3–10], but new research topics are meanwhile identified. In fact, up to six new research areas have now been detected from the obtained 13 topics. For instance, specific topics like evaluation of effectiveness and benefits of ITS, mobile and Table 3 Continued TOPIC 13 propagation 0.019 – – – – – – congestion 0.018 – – – – – – estimated 0.017 – – – – – – rights 0.017 – – – – – – IET Intell. Transp. Syst., 2010, Vol. 4, Iss. 3, pp. 201–211 207 doi: 10.1049/iet-its.2009.0102 &The Institution of Engineering and Technology 2010 www.ietdl.org wireless communications in ITS or models for traffic demand forecasting have not been previously established. They have emerged as the result of the incorporation of new technologies, and also due to the necessity of modelling and assessing their social and economical impact. The proposed categorisation should help researchers and practitioners to clarify their position for future work because the obtained 13 topics represent the most recent issues and emergent technologies in the ITS world. The topic correlation matrix (Table 4) shows that topics are poorly correlated with each other, which means that topics are well defined and their scope is clearly delimited. Nevertheless, it is impossible to achieve perfectly delimited topics with independent scopes. There is always some degree of overlap. Overlapping can be used to obtain several major paradigms attending to the topic similarity. For this purpose, a multivariate statistical technique like multidimensional scaling was used [33]. This analysis consists of projecting the works on a two-dimensional map, using the data from the correlation matrix as input data. Fig. 3 shows the obtained map. The obtained RSQ coefficient (0.93430) and Kruskal’s stress (0.15374) suggest that goodness of fit is very acceptable (the RSQ coefficient is the squared correlation index R 2 that measures the model fit to the data, and its minimum acceptable score is 0.6 Figure 3 Multidimensional scaling Table 4 Topic correlation matrix T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T1 1.00 0.13 0.10 0.07 20.05 20.03 0.06 0.14 0.07 20.01 20.10 0.03 0.06 T2 0.13 1.00 0.01 0.00 0.12 20.07 0.16 0.07 0.16 20.07 0.03 0.10 0.04 T3 0.10 0.01 1.00 0.03 0.13 0.01 20.06 0.10 0.07 0.13 20.02 20.02 0.08 T4 0.07 0.00 0.03 1.00 20.05 0.14 20.05 0.19 20.02 0.05 20.02 0.21 0.21 T5 20.05 0.12 0.13 20.05 1.00 0.05 0.21 20.02 0.28 0.05 20.03 20.09 20.08 T6 20.03 20.07 0.01 0.14 0.05 1.00 20.04 0.04 0.01 0.03 0.34 20.02 0.00 T7 0.06 0.16 20.06 20.05 0.21 20.04 1.00 20.06 0.12 20.10 20.06 20.02 20.07 T8 0.14 0.07 0.10 0.19 20.02 0.04 20.06 1.00 20.01 0.08 0.03 0.25 0.15 T9 0.07 0.16 0.07 20.02 0.28 0.01 0.12 20.01 1.00 20.01 0.04 20.03 0.02 T10 20.01 20.07 0.13 0.05 0.05 0.03 20.10 0.08 20.01 1.00 20.04 0.00 20.10 T11 20.10 0.03 20.02 20.02 20.03 0.34 20.06 0.03 0.04 20.04 1.00 0.04 0.13 T12 0.03 0.10 20.02 0.21 20.09 20.02 20.02 0.25 20.03 0.00 0.04 1.00 0.30 T13 0.06 0.04 0.08 0.21 20.08 0.00 20.07 0.15 0.02 20.10 0.13 0.30 1.00 208 IET Intell. Transp. Syst., 2010, Vol. 4, Iss. 3, pp. 201–211 &The Institution of Engineering and Technology 2010 doi: 10.1049/iet-its.2009.0102 www.ietdl.org [34]) and that the map exhibits a good approximation of reality. The proximities of topics on the map show their similarity and suggest a meaning for the axes. The horizontal axis is related to the application area. Topics located to the left part of the map (T4, T8, T12 and T13) are focused on the improvement of traffic in urban area. Topics located on the centre of the map consider ITS tools from two different perspectives, the implementation details represented by topics T3, T6, T10 and T11 at the top of the map and their benefits and potential use represented by T1 and T2 at the bottom of the map. Finally, topics located to the right part of the map are focused on vehicle and traffic management (T5, T6 and T7). The vertical axis is related to the level of hardware implementation, as it can be clearly deduced from topics in the upper half of the map, concerned with the implementation details of ITS tools (T3, T6, T10 and T11) or the mobile and wireless communication possibilities supporting traffic management (T5). Comparing these results with previous classifications detailed in Section 2, notice that ‘Benefits and potential use of ITS tools’ were not considered from a technological perspective and the ‘implementation details of ITS tools’ were not considered from a market area perspective. Consequently, the map of Fig. 3 summarises a more complete vision of ITS. 5 Conclusion The main contribution of this paper is a global view of ITS, which could be used by future researchers as a state of the art of methods, techniques and application areas. The analysis is intended to offer new perspectives into what is viewed as important to build upon, providing valuable insights into both what research is important and where the field of ITS is heading. The proposed methodology for producing this global view is based on the semantic analysis of keywords and abstracts of papers indexed by the ISI. As a difference to author or journal co-citation methods, semantic analysis is focused on the content of papers, so results are not biased by cites to irrelevant literature or recurrent cited papers. A total of 1147 papers have been analysed covering a large variety of topics related to ITS, including the most recent issues and emergent technologies. As a result of the analysis, 13 paradigms were obtained. Using a multidimensional scaling, they have been represented on a bidimensional map, illustrating the most related paradigms as well as the bridges among them. Furthermore, this study also defines a starting point for other analyses aimed at a better understanding of the ITS field. This continuous analysis is considered necessary as ITS is an evolutionary field influenced by changes in technology, with changing services and support to end users. 6 Acknowledgments The authors gratefully acknowledge support provided by the Spanish Ministry of Education and Science (project with reference DPI2007-60128) and the Consejerı ´ade Innovacio ´n, Ciencia y Empresa (Research Project with reference P07-TIC-02621). 7 References [1] ANDRISANO O.,VERDONE R.,NAKAGAWA M.: ‘Intelligent transportation systems: the role of third generation mobile radio networks’, IEEE Commun. Mag., 2000, 38, (9), pp. 144–151 [2] TORAL S.,VARGAS M.,BARRERO F.: ‘Embedded multimedia processors for road-traffic parameter estimation’, Computer, 2009, 42, (12), pp. 61–68 [3] MCQUEEN B.,MCQUEEN J.: ‘Intelligent transportation systems architectures’ (Artech House, 1999) [4] NATVIG M.K.,WESTERHEIM H.: ‘National multimodal travel information – a strategy based on stakeholder involvement and intelligent transportation system architecture’, IET Intell. Transp. Syst., 2007, 1,(2), pp. 102–109 [5] TORAL S.,VARGAS M.,BARRERO F.,ORTEGA M.G.:‘Improved sigma-delta background estimation for vehicle detection’, Electron. Lett., 2009, 45, (1), pp. 32–34 [6] FIGUEIREDO L.,JESUS I.,MACHADO J.A.T.,FERREIRA J.R.,MARTINS DE CARVALHO J.L.: ‘Towards the development of intelligent transportation systems’. Proc. 2001 IEEE Intelligent Transportation Systems, Oakland, CA, USA, 2001, pp. 1206–1211 [7] WANG F.-Y.,MIRCHANDANI P.B.,ZHENG N.:‘Advancesand trends in research and development of intelligent transportation systems: an introduction to the special issue’, IEEE Trans. Intell. Transp. Syst., 2004, 5,(4), pp. 222–223 [8] ZHENG N.,KAWASHIMA H.: ‘Advances and trends in research and development of vehicular electronics and safety: special section’, IEEE Trans. Intell. Transp. Syst., 2007, 8, (1), pp. 106–107 [9] Artemis Strategic Research Agenda Working Group: ‘Strategic research agenda’ (Ideo Ltd, 2006, 1st edn. edn.) [10] US Department of Transportation, Federal Highway Administration: ‘Intelligent transportation systems benefits, costs and lessons learned’ (Mitretek Systems, 2005) IET Intell. Transp. Syst., 2010, Vol. 4, Iss. 3, pp. 201–211 209 doi: 10.1049/iet-its.2009.0102 &The Institution of Engineering and Technology 2010 www.ietdl.org