International Transportation (71) 1 | 2019 38 Advanced automation in railway operations Impacts, requirements and potentials Automation, Energy consumption, Railway, Simulation Automation is already present in many areas of the railway sector. However, to achieve set climate goals, increase capacity, reduce costs and offer an attractive transport service, it is essential to systematically apply ATO (Automatic Train Operation) or higher Grades of Automation (GoA). This paper summarises the findings of a study regarding the impacts, requirements and potentials of higher automation in the railway sector. The analysis distinguishes between (i) mainlines and branch lines as well as (ii) passenger transport, freight and mixed traffic. Furthermore, results based on a model simulating energy consumption highlight the importance of energy-efficient driving. Martina Zeiner, Martin Smoliner A TO is considered a subsystem with different functions depending on the GoA (as defined by the UITP) and must be combined with ATP (Automatic Train Protection) to ensure safety. ATP together with DAS (Driver Advisory Systems) are classified as GoA1 and are already used by railways. GoA2 combines ATP and ATO; where ATO executes traction and brake commands. Much effort is currently put in field trials for GoA2, albeit one can also find existing examples of GoA2, such as the Thameslink project in London. GoA4 corre- New trends in transport systems For the 14 th consecutive time the European Platform of Transport Sciences - EPTS - awards a prize dedicated to young transport researchers. The prize is named “European Friedrich-List-Prize” to honour the extraordinary contributions of Friedrich List, the visionary of transport in Europe of the 19 th century, being a distinguished economist and respected transport scientist committed to the European idea. The European Friedrich-List-Prize is awarded for out-standing scientific papers in each of the categories Doctorate paper and Diploma paper. The submitted papers address topics in the transport field within a European context and from a European perspective. In 2019 around 150 scientific works have been nominated and evaluated. The award will be conferred during the 17th European Transport Congress in Bratislava (Slovakia) on 13th June 2019, and the results will be introduced on the website www.international-transportation.com. In the following you can find a small random selection of this year’s submissions summarized in drafts. SCIENCE & RESEARCH European Friedrich List Award Bratislava Photo: Džoko Stach on Pixabay European Friedrich List Award SCIENCE & RESEARCH International Transportation (71) 1 | 2019 39 sponds to fully automatically run vehicles and until now has only been applied in urban metro lines. Requirements for higher automation in the railway sector are listed in figure 1. Since ATP is a safety requirement as of GoA2 and to ensure interoperability, many institutions and suppliers support the idea of ATO over ETCS. Efforts are currently underway to incorporate new specifications for GoA1 and 2 in the TSI [2]. Furthermore, adaptions in national legislation, liability issues (of trial runs), certification issues and harmonised authorization processes have to be considered and solved. To ensure the safe guidance of a train a continuous ATP must be implemented and continuous information, usually known to the driver, needs to be submitted to the ATO. In Europe ETCS Level 2 is regarded as the basis for ATO. However, the current infrastructure and slow migration process of ETCS makes the use of a harmonised, sophisticated ATP very unrealistic. ATP solutions based on satellites should thus be examined together with migration concepts in case of ATP other than ETCS-[3]. In order to increase energy efficiency and punctuality ATO has to be combined with DAS (providing an optimised speed profile for one train). To optimise train movements in a whole network, ATO must be connected to a cross-network Traffic Management System (TMS). This would mean adapting trajectories to the current state of the traffic continuously to avoid unnecessary stops, reactionary delays or conflicts. One approach is known as dynamic capacity optimisation; it is based on an automatically computed timetable in real-time combined with ATO and can reduce headways (90 to 100-sec.)-[4]. Technical equipment on wayside and trainborne level will need to be adjusted depending on the GoA. As of GoA3, the train has to take over the driver’s visual functions. For wayside obstacle detection, solutions stem from drone-based cameras to fibre optic sensing. The installation of laser or radar sensors combined with image processing at level crossings or fences at platforms are conceivable solutions [5]. As for on board obstacle detection the combined installation of radar, infrared, laser or cameras is suggested because of different characteristics in reach and dependence of weather-[6]. Different systems can benefit from increasing automation according to their boundary conditions. Capacity problems are particularly prevalent in passenger transport, especially on mainlines. Solutions as of GoA2 in connection with TMS show great potential in passenger transport and mixed traffic for coping with peak demand in hubs [4]. The need for additional infrastructure (as of GoA3) could therefore be replaced by means of a dispositive level. Comfort can already be achieved as of GoA2, since ATO can balance e.g. aggressive styles of driving. The use of TMS reduces waiting time, increases reliability and punctuality which has added value for freight and passenger transport. In a first step this can already be achieved to a certain degree with DAS. Introducing TMS plus fully automatically run vehicles on branch lines could bring about a cost-effective and demand-based transport service. There is a common understanding that safety increases by taking out the human factor. However, risks caused by new technologies as of GoA3 must be less in character than the human-risk factor in terms of cyber security, failures of providers, manufacturers or systems. Recently developed “intelligent” vehicles (equipped with a centre buffer coupling and being able to perform an automatic brake test) could replace the remaining manual work of coupling processes [7]. Safety in shunting could thus be increased considering the high risk of accidents. A useful way to save energy is to exploit the acceleration, cruising, coasting and braking phase more energyefficiently. An energy calculation program [8] was elaborated to depict different driving behaviours and estimate the respective energy consumption. The model is based on the total train resistance which occurs along a train journey on a random route and can be expressed in energy needed for that section. Energy consumed by auxiliary functions is also considered. Different driving styles were computed for a section on one Austrian mainline taking into account five train types (passenger and freight); energy recovery remained unconsidered. Figure 2 displays three speed profiles for a local regional train. Case maximum top speed demonstrates a tight speed profile which might counteract a reactionary delay. Energy-efficient driving is shown by limiting the top speed and yet arriving on time (considering 5 % Fully automatic Train-Operation Operational- requirements Operational-rules and regulation Traffic-Management-(Dispatching,-TMS) Operation-according to demand Safety &-Security Certification Authorisation Tests-&-Validation Liability Vehicle / -Infrastructure Localisation / -Positioning Communication Obstacle Detection ATO-over ETCS Interoperability Legal-and normative- framework Technical- requirements Figure 1: Requirements for automation according to [1] International Transportation (71) 1 | 2019 40 buffer time). Case 3 represents the same train without any stops. The ratio of the train resistances shows that the major part of consumed energy can be led back to acceleration resistance. A top speed reduction can save up to around 50 % of energy compared to a tight speed profile. The decrease in energy consumption generates a 35 % higher running time which highlights that the degree of energy reduction is not wedded to the degree of increase in travel time. To a certain extent the results must be treated with caution, since in reality a train will not accelerate up to the permitted top speed every time before a stop. It can be assumed that the energy savings will be lower; better corresponding to what can be found in the literature, e.g. [9]. Furthermore, the energy-efficient driving profiles only concern the optimisation of a single train. The optimisation of several trains might lead to less reduction for the single train. An appropriate TMS is therefore necessary to save energy on a networkwide level. Investigations on freight trains confirm the need of an improved management of traffic. Around 35 % of energy could be saved on that route without putting freight trains aside. Energy-efficient driving could reduce energy costs by 10 % on average for one train in one year [10]. This could in particular increase the competitiveness of freight traffic. Whereas cost cuts by replacing drivers is a doublesided issue, the economic benefit in shunting is certain due to the possible decrease in manual labour. The outcome of the study underlines the importance of elaborating automation on a dispositive level (at best cross-network TMS) if energy and capacity improvement are to be achieved. Hence, punctuality along with cost-effective offers bring added value to the customer and boost the railway sector. In case of passenger and mixed traffic, energy savings and capacity increases can already be achieved with GoA1 and 2. The results of different driving behaviours underpin the benefits of DAS and ATO. One potential in freight transport can be exploited with “intelligent” freight trains improving safety and reducing costs. Nevertheless, one must consider side effects of higher levels of automation in terms of technical requirements as well as in the context of legal aspects. Moreover, there are limits to operational optimisation; results of the energy calculation show that auxiliary functions consume one sixth of the total energy consumption. Opportunities for more environmentallyfriendly vehicles should thus be contemplated. Finally, it should be noted that the study only focuses on electrified railways. On non-electrified lines the focus should also be placed on alternative traction technologies. ■ REFERENCES [1] M. Meyer zu Hörste, Fully automatic railway operation: Concept and Conditions, Vienna, 28.11.2017 [2] R. Treydel, Development of the harmonised European specifications for mainline ATO, Vienna, 28.11.2017 [3] J. Trinckauf, “Der Bahnbetrieb auf dem Weg zur Digitalisierung und Automatisierung,” Deine Bahn, no. 09, pp. 7-9, 2017 [4] U. Weidmann et al., “Dynamische Kapazitätsoptimierung durch Automatisierung des Bahnbetriebs,” Eisenbahn-Revue, no. 12, 606-611, 2014 [5] J. Pachl, “Betriebliche Randbedingungen für autonomes Fahren auf der Schiene,” Deine Bahn, no. 09, pp. 11-19, 2017 [6] O. Gebauer et al., “Autonomously driving trains on open tracks— concepts, system architecture and implementation aspects,” it - Information Technology, vol. 54, no. 06, pp. 266-279, 2012 [7] B. Müller-Hildebrand, “Automatisierung und Digitalisierung im Schienengüterverkehr,” ZEVrail, no. 08, pp. 301-303, 2017 [8] M. Messner, Berechnung des Energieverbrauchs für Triebfahrzeuge, TU Graz, 17.05.2014 [9] M. Marinov, ed., Sustainable Rail Transport: Proceedings of RailNewcastle Talks 2016, Springer International Publishing, 2018 [10] J. Winter et al., “Fahrerassistenz-System,” Signal + Draht, no. 10, pp. 6-14, 2009 Martina Zeiner, Dipl.-Ing. BSc University Assistant, Institute of Railway Engineering and Transport Economy, Graz University of Technology (TU Graz), Graz (AT) martina.zeiner@tugraz.at Martin Smoliner, Dipl.-Ing. BA BSc MA University Assistant, Institute of Railway Engineering and Transport Economy, Graz University of Technology (TU Graz), Graz (AT) martin.smoliner@tugraz.at 0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500 max.-top speed top-speed reduction no-stops 0 time-[min] Energy-consumption-[kWh] Energy-consumption-due-to-train-resistance-and- auxiliary-functions-for-different-driving-styles Auxiliary-Functions Starting-Resistance Acceleration-Resistance Resistance-according-to-alignment Running-Resistance calculated-running-time max.-allowed-running-time max. top---------top-speed-------no-stops speed------------reduction 0 160 0 50 Velocity--[km/ h] Distance-[km] Speed-profiles-for-different-driving-styles permitted-top-speed max.-top-speed top-speed-reduction no-stops Figure 2: Results of the simulation model SCIENCE & RESEARCH European Friedrich-List-Prize European Friedrich List Award SCIENCE & RESEARCH International Transportation (71) 1 | 2019 41 Challenging assumptions about traveller behaviour The benefits and challenges of using Bluetooth data to examine repeated behaviour Big data, Travel behaviour, Variability, Bluetooth data Emerging data sources provide new opportunities to test how well long held assumptions in transportation reflect reality. This article presents a case study which uses one year of data from 23 fixed Bluetooth detectors to examine the regularity of individual travel behaviour over time. New insights were obtained into the relationship between spatial and time of day variability and the proportion of travellers with very regular travel patterns. This type of research is challenging, however, due to the large amounts of data involved and the need to develop new methods to analyse the data. Fiona Crawford S implifying assumptions are a necessary part of the transport modelling process. Without such assumptions we would be overwhelmed by the complexity of people’s travel behaviour and the countless types of variability present. In some cases, long standing assumptions which were originally devised when data was scarce have become so ingrained that they often go unspoken and unchallenged. As the amount of data available about Figure 1: Map of Bluetooth detector locations in the Wigan area (from-[1]) SCIENCE & RESEARCH European Friedrich List Award International Transportation (71) 1 | 2019 42 User class Subclass Number of devices Number of trips Average trip frequency Spatial variability Time of day variability Frequent travellers F1 16,634 1,144,115 1-2 per week More variable More variable F2 8,163 820,221 1-2 per week Less variable Less variable F3 3,089 815,504 5 per week Less variable Less variable F4 5,809 1,590,437 5 per week More variable More variable Very frequent travellers V1 1,901 1,302,874 2 per day Highly variable Average of F3 and F4 V2 195 349,002 5 per day Highly variable Average of F3 and F4 Table 1: Summary table of frequent and very frequent travellers in the Bluetooth case study both the transport network and traveller behaviour grows rapidly, so do the opportunities to examine the validity of different assumptions. Individuals’ travel behaviour is often assumed to be highly repetitive, particularly during the peak period, with the same people being observed on the roads each day. Such assumptions are increasingly questionable given changes in working hours and the locations in which work is undertaken. Newly emerging types of data open up new opportunities to examine people’s repeated trip behaviour to determine how repetitive their travel choices are. This article describes research focusing on the regularity of individuals’ behaviour using data from fixed Bluetooth sensors, which are relatively cheap to purchase and can be permanently installed to collect data for large numbers of travellers over long periods of time. A case study using Bluetooth data Bluetooth is a wireless technology which is used in a vast array of personal devices, including mobile phones and fitness trackers. The technology is also commonly used in vehicles, for example to enable hands-free calls or to connect devices to in-vehicle sound systems. A key aspect which makes the technology useful for transport planners is that each device has a unique identifier, known as a MAC address, which can be captured by fixed Bluetooth detectors. A number of cities around the world have successfully used the technology for measuring travel times by matching devices between detector locations using their MAC addresses. The potential to match MAC addresses between days has not been explored previously, however. This research examined the different groups of road users (in motorised vehicles) which exist in a case study area, based on their repeated trip behaviour. The objective was to assess the assumption that most traffic is the result of people travelling to the same place at the same time each day. The case study is within Greater Manchester in the north of England, where over 500 fixed Bluetooth sensors have been installed alongside the road network for the purpose of monitoring travel times. This research focuses on 23 detectors in and around the town of Wigan, as shown in figure 1. Data from 2015-01-01 to 2015-12-31 was analysed. Processing and analysing the data was challenging. Firstly, consecutive observations for the same device needed to be identified and then filtered so that only observations which were likely to relate to a direct trip between the two Bluetooth detectors in a motorised vehicle were retained. This process involved examining the free flow travel times between each pair of sensors, but also the travel times recorded by surrounding Bluetooth devices (see [1] for more details). Determining the best way to examine repeated trip behaviour using the data was also challenging. The data is relatively similar to the data obtained from Automatic Number Plate Recognition (ANPR) systems, but the scale of data (including the time period over which unique identifiers could be matched and the number of detectors) was relatively large. Bluetooth data also has the added complication that not all Bluetooth-enabled devices will be detected when passing a sensor and therefore one missing detection within a trip through the town is more common than it would be for ANPR data. It was straightforward to calculate a lower bound for the number of trips made by each Bluetooth device during the year. The number of trips is an underestimate of the trips made by the device as not all Bluetooth enabled devices passing a sensor will be detected, and devices can only be detected if they are switched on and the Bluetooth functionality is enabled. It may also be an underestimate of the number of trips made by the owner of the device, as the device may not be taken on all trips. Measuring the spatial and time of day variability for each device, however, was much more challenging. In both cases, new methodologies were required. For spatial variability, the distribution of each traveller’s trips across different ‘spatial sets’ was measured. These ‘spatial sets’ were defined by grouping together similar trips (denoted by the list of sensor locations at which a device was detected). The grouping used a method called Sequence Alignment to compare trips based on their components (the sensors where detections were made) and the order in which they appear. Sequence Alignment is used in bioinformatics to compare protein sequences and has also been used in the social sciences [2]. Time of day variability was measured for each traveller by identifying clusters of the times of day they are observed at their most common sensor location and then examining the number of clusters and the average variance of the clusters. After producing measures of trip frequency, spatial variability and time of day variability for each traveller, different user classes could be identified within the data. Of the 7,480,204 trips which were recorded by Bluetooth detectors during the year, 81 % were made by devices European Friedrich List Award SCIENCE & RESEARCH International Transportation (71) 1 | 2019 43 which were classified as frequent or very frequent travellers during the analysis process. The six subclasses of Bluetooth devices in these categories are shown in table-1. The frequent travellers can be separated into two groups based on trip frequency (F1 and F2 versus F3 and F4). Each of these groups contains one subclass which is larger and has higher levels of both spatial and time of day variability for individual travellers. Even when considering just 23 sensors around a relatively compact town centre, there appears to be many travellers who are not highly regular in their trip making. The very frequent travellers are also of interest as 22 % of the trips observed in the Bluetooth data are made by this group. These travellers have similar time of day variability to F3 and F4, but they have more spatial variability. These travellers make trips in more of the different ‘spatial sets’ than frequent travellers, but as they make so many more trips during the year, these very frequent travellers also repeat trips more often. Given that the estimated number of trips is a lower bound on the actual number of trips made, it is possible that these devices are not recording personal travel, but travel related to business activities. Bluetooth data therefore provides a different perspective from travel diaries which typically collect data on personal travel only. Challenges and opportunities This small case study has demonstrated that new insights, in this case into repeated trip behaviour, can be obtained from newly emerging data sources. In this example, Bluetooth data was used. The substantial amount of research which has been undertaken to examine Bluetooth data’s usefulness for measuring travel times (including [3] and [4]) provides some confidence in the data, but different kinds of research are required if the data is being considered for a different type of application. For example, more research is required into how people use the wide range of Bluetooth-enabled devices now available, such as fitness trackers and in-vehicle audio systems. Trust in the data is only part of the challenge, however. Processing large quantities of data can be time consuming and may require large amounts of computing power. New analytical techniques may be required to gain insights from the data, perhaps from fields such as data science. Therefore, whilst more and more opportunities are opening up to allow us to challenge long held assumptions about travel behaviour, it is essential that we ensure, as a community, that we are developing the necessary skills and methods and ensuring that we have access to the resources we need to make the most of the opportunities ahead. ■ The data used in this research was provided by Transport for Greater Manchester REFERENCES [1] Crawford, F., Watling, D.P. and Connors, R.D. (2018), Identifying road user classes based on repeated trip behaviour using Bluetooth data. Transportation Research Part A: Policy and Practice. 113 pp. 55-74. doi: 10.1016/ j.tra.2018.03.027 [2] Crawford, F. (2017), Methods for analysing emerging data sources to understand variability in traveller behaviour on the road network. University of Leeds. Available at: http: / / etheses.whiterose.ac.uk/ 18758/ 1/ Crawford_F_ITS_PhD_2017.pdf [3] Quayle, S.M., Koonce, P., Depencier, D. and Bullock, D.M. (2010), Arterial Performance Measures with Media Access Control Readers: Portland, Oregon, Pilot Study. Transportation Research Record. 2192 (1), pp. 185-193 [4] Araghi, B.N., Olesen, J.H., Krishnan, R., Christensen, L.T. and Lahrmann, H. (2015), Reliability of Bluetooth Technology for Travel Time Estimation. Journal of Intelligent Transportation Systems. 19 (3), pp. 240-255 Fiona Crawford Research Fellow in Transport Studies, Centre for Transport and Society, University of the West of England, Bristol (GB) fiona.crawford@uwe.ac.uk Risk analysis of dangerous goods transportation Risk analysis, Dangerous goods, Human resources, Risk mitigation The paper deals with transport of dangerous goods by road (ADR). Main contribution is the development of the algorithm for evaluation and management of human factor risks in the field of dangerous goods transport. There is presented a systematized approach and unambiguously structured the gradual use of qualitative, quantitative and semiquantitative methods for the risk assessment. Following methods are used: Check-list; What, if; Failure Modes Effects and Causes Analysis (FMECA); Human Reliability Assessment (HRA), Fault Tree Analysis (FTA). Libor Krejčí T ransportation of dangerous goods takes place as an important part of European economy. Due to the nature of dangerous goods (e.g. chemicals, petroleum products, explosives), such substances are causing significant threat to road or railway users, inhabitants, infrastructure and the environment. A certain degree of risk is at any stage of dangerous goods manipulation, during its production, storage, han- SCIENCE & RESEARCH European Friedrich List Award International Transportation (71) 1 | 2019 44 Figure 1: Flow chart for the risk analysis of dangerous goods transportation dling, transportation and consumption. Therefore the international legislation standards set a legal framework for dangerous goods transportation in Europe. Dangerous goods transportation by road is regulated by the Agreement concerning the international carriage of Dangerous Goods by Road (ADR) [1]. Even though there are the regulations in this area focused on training of drivers, dangerous goods safety advisors and other professionals in the supply chain, still the human factor plays a major role, causing the most accidents in this transport sector. Qualitative, quantitative and semi-quantitative methods for the risk analysis and assessment had been employed to address an unacceptable risk. Author determined the technological process for the application of the individual methods in gradual steps, structured the evaluated areas, set the evaluation criteria and set the unacceptable risk limits for the individual quantitative methods. Methods used Check-list - the method was used to identify all the major causes of previous accidents related to human factor in the dangerous goods transportation. All the identified hypothetical causes of accidents related to handling of dangerous goods were included into a risk register and then compared with relevant international databases of past accidents Major Accident Reporting System (eMARS) [2] and National Transportation Safety Board (NTSB) [3]. Risk areas with a reference to any accident in the databases were further analysed with more systematised follow up methods. What, if - the method enables to identify risks across the broader spectrum of the dangerous goods transportation processes. A group element of the method was used to stimulate the cooperation and creation of various scenarios. A multidisciplinary risk analysis team was set up for this purpose, consisting of risk manager, driver transporting dangerous goods in packages, driver transporting dangerous goods in tank vehicle, warehouse worker handling dangerous goods, dangerous goods safety advisor, policeman and fireman. Failure Modes Effects and Causes Analysis (FMECA) - the group inductive method for assessing probability of occurrence, impact and detection of undesired events. For the purposes of the analysis, the system was defined as a transportation process of dangerous goods by road and the failure a road traffic accident, which could lead to the leakage of dangerous goods. The FMECA analysis extends the (Failure Mode and Effects Analysis) FMEA analysis by classification each type of failure. Human Reliability Assessment (HRA) - the method was used to target the impact of human activity on the functionality of the dangerous goods transportation system. The qualitative and quantitative forms of the method were used. The qualitative form to identify the possibility of human errors and their causes in order to reduce the probability of errors. The quantitative form of the method provided input data on human failure into the follow-up analysis. Fault Tree Analysis (FTA) - the method was used primarily for a highly systematic approach, allowing sufficient flexibility for the analysis of different factors at the same time. The undesirable peak event was a road traffic accident, which could lead to the leakage of dangerous goods. Causal factors were partly utilized from the previous methods applied. Other factors induced by this method were included into the follow-up analysis as well. At the final stage of the risk assessment, the calculation of the individual and social risk for the population during the transportation of dangerous goods was done using the F/ N curves (frequency “F” where a certain number of “N” consequences may occur). Finally, determination of the acceptability of individual and social risk caused by dangerous goods transportation, during handling and along the road was provided. In order to asses risk mitigating measures the determination of the ALARP (As Low As Reasonably Practicable) zone was utilised. Gradual implementation of the methods for risk identification, impact analysis and risk assessment is illustrated in figure 1. Conclusions All the aforementioned methods had been applied in a number of fields of human activities (engineering, petrochemical, industrial), where the continuous risk control is important. However, utilisation of these methods has Start risk analylis of dangerous goods transportation 2 Current scienti c approach 3 Accident databases 4 Scienti c methods 1 State of the art 5 Check-list 6 What, if 7 FMECA 8 HRA 9 FTA 10 Acceptable risk? 11 Risk mitigating measures 12 Prevention of accidents 13 Mitigation of accident consequences End - acceptable risk + - European Friedrich List Award SCIENCE & RESEARCH International Transportation (71) 1 | 2019 45 been rather rare and isolated in the area of dangerous goods transportation. The developed algorithm for evaluation and management of human factor risks in the area of dangerous goods transportation will help to reduce the risk caused by the most risky human factor. Implementation by the carriers, consignors (loader, packer, filler, tank-container/ portable tank operator) and consignees (unloader) of dangerous goods, minimizes the potential negative impacts and contributes to the overall safety improvement in Pan-European region. It is always necessary to implement the algorithm in full extent as designed by the author. It is not appropriate to implement the individual methods individually or independently of each other, in order to obtain only certain partial knowledge. There are mutual relationships among the methods included into the algorithm. These methods complement each other and allow approaching the risk area from an aggregated viewpoint to individual sub-processes. Only with the systematic gradual implementation of individual methods, it is possible to gain the full advantage of their distinctions. The author prepared the set of measures to address the risk of dangerous goods transportation by road. Even though these measures are beyond the scope of this article, all the measures are aimed at prevention of accidents or mitigation of their potential negative impacts on population, infrastructure and the environment. Exploitation of the common developed algorithm addressing the risks in the whole dangerous goods supply chain is going to be facilitated especially by the dangerous goods safety advisors, who are the main drivers of safety and security in target companies. As a consequence risk mitigating measures shall be adopted by other professionals in the dangerous goods transportation. ■ REFERENCES [1] United Nations Economic Commission for Europe (2019), European Agreement concerning the International Carriage of Dangerous Goods by Road (ADR), https: / / www.unece. org/ trans/ danger/ publi/ adr/ adr2019/ 19contentse.html [2] European Commission (2019), eMARS (the Major Accident Reporting System), https: / / emars.jrc.ec.europa.eu/ [3] National Transportation Safety Board (2019), Hazardous Materials Accidents Reports, http: / / www.ntsb.gov/ investigations/ AccidentReports/ Pages/ hazardous.aspx Libor Krejčí, Ph.D. CDV - Transport Research Centre, Brno (CZ) libor.krejci@cdv.cz TSCLab - Traffic Signal Control Laboratory A tool for performance monitoring and evaluation of adaptive traffic signal control in VISSIM TSCLab, VISSIM, Signal control, Measures of effectiveness Adaptive Traffic Control Systems (ATCS) have been widely implemented for urban traffic control due to their capability to alleviate congestion. The evaluation of the effectiveness of complex ATCS is challenging and presents an open problem. The most important issue is to identify whether the ATCS fulfills the goals envisioned to be achieved. In this paper, development of TSCLab (Traffic Signal Control Laboratory), a MATLAB based tool for evaluation of ATCS is presented. To proof the capabilities of TSCLab, the effectiveness of the UTOPIA/ SPOT ATCS as the use case has been evaluated. Daniel Pavleski P rior implementation of an ATCS in an urban environment, it has to be evaluated using realistic traffic scenarios. This is important in order to assess the possible improvement of LoS and to analyze the cost-benefit ratio before a costly upgrade of the transport infrastructure [1]. Mostly used approach for this is software-in-the-loop where a microscopic traffic simulator in combination with an ATCS is applied [2]. Many researchers address this problem in order to implement an appropriate framework, ensure realistic traffic scenarios from different world regions, enable in-depth behavior analysis of the managed urban transport network and define appropriate MoEs. This paper also tackles the problem of evaluating ATCS using the software-in-the-loop approach. SCIENCE & RESEARCH European Friedrich List Award International Transportation (71) 1 | 2019 46 Figure 1: Diagram for monitoring of the signal time changes and vehicle arrivals [6] Figure 2: Table with results related to progression [6] ATCS performance selection In the absence of information for the performance of adaptive traffic signal control, the quality of signal operations cannot be determined, and the functionality of the control strategies cannot be validated [3]. Therefore, the monitoring of these so-called “live” systems i.e. the monitoring of their real-time “responses” to certain traffic state and a specific objective function is crucial. Due to the complexity of ATCS, the process of evaluating their effectiveness requires using of measures of effectiveness with in-depth insights into the traffic situations of the controlled signalized intersection. The available literature describes a variety of MoEs that can be used for ATCS performance analysis [3, 4, 5]. In the report [3], are outlined candidate Measures of Effectiveness (MoE) for each operational objective and each MoE is denoted as a candidate since it is not necessary to calculate or compare all of the measures to validate the functionality of a system. The report [4] documented an extensive portfolio of performance measures for evaluating traffic signal systems with emphasis on performance measures obtained from high-resolution data and from external travel time measurements. The report [5] presents the next step toward integrating traffic signal performances measures in traffic signal systems facilitated by high-resolution controller event data. The mentioned reports have been used as a starting point and base for selection of MoEs for ATCS performance evaluation. Selected measures that have been used for ATCS performance evaluation in this paper are: • Cycle length and Green time duration; • Maximum green time utilization ratio; • Arrived vehicles per cycle; • Served vehicles per green signal; • Green/ Red occupancy ratio; • Queue length; • Delay; • Stops; • Percent of arrived vehicles on green signal; • Platoon ration & Arrival type. Development of TSCLab - Traffic Signal Control Laboratory In order to apply the selected measures for performance evaluation of ATSC which are not featured in VISSIM, a MATLAB based tool named TSCLab (Traffic Signal Control Laboratory) with graphical user interface was developed and connected to VISSIM trough COM interface of VISSIM. This tool can access, gather and visualize relevant data generated by VISSIM, needed for calculation of selected measures for performance evaluation of ATSC in VISSIM based microscopic simulation environment [6]. The main graphical user interface of the TSCLab tool is consists of three main sections. The first section refers to measurements assignment where objects defined in the VISSIM model such as Nodes, Data collection Measurements, and Queue Counters, can be specified. The second section refers to traffic signal control assignment where objects defined in VISSIM model such as Traffic Light Controller, Signal Groups, Detector ports, and Detectors, can be specified. The third section is related to simulation where parameters for simulation in a VIS- SIM model such as Start/ End of Time Period and Resolution can be specified. TSCLab enables diagrams for real time monitoring of: • Signal Times/ Detector occupancy; • Signal Times/ Vehicle arrivals; • Max green time utilization ratio; • Vehicle arrivals per cycle; • Served vehicles per green time; • Green/ Red occupancy ratio; • Vehicle queue link profile; • Percent of vehicle arrivals on green time; and • Vehicle Platoon ratio & Arrival type. An example of diagram for real time monitoring is shown in figure 1. In addition, TSCLab enables tables with outputs that refer to: (i) Signal timing; (ii) Throughput; (iii) Capacity and (iv) Progression. An example of output data related to signal timing is shown in figure 2. Evaluation of ATSC - a case study To proof the capabilities of the developed TSCLab tool it has been applied to evaluate the effectiveness of the UTOPIA/ SPOT ATCS using an isolated signalized urban intersection as case study. A VISSIM model for urban network of seven signalized intersections located in wider central area of Skopje was developed and the Intersection denoted as I2, was selected for testing. All signalized intersections in the study area are managed by the ATCS UTOPIA. Therefore, UTOPIA as a ”black box” was connected to the developed VISSIM model through the UTOPIA VISSIM Adapter (UVA). With this connection, UTOPIA manages the traffic signals for the simulated road network and VISSIM provides the needed European Friedrich List Award SCIENCE & RESEARCH International Transportation (71) 1 | 2019 47 traffic data from the sensors. Both, traffic signal commands and sensor measurements are refreshed every second [6]. In order to evaluate the performances of ATCS UTO- PIA, the morning peak hour (7: 15 to 8: 15) in a typical working day was chosen for analysis. To create realistic simulation model in VISSIM the following information have been obtained: 1) Network layout; 2) Familiarity with site operation and driver behaviour; 3) Traffic flows and turning proportions; 4) Traffic flow compositions; 5) Bus frequencies; 6) Bus stop locations; 7) Bus stop dwell times; 8) Signal timings and controller logic; 9) Saturation flow; 10) Queue lengths; and 11) Mandatory speed limits. A simulation period of 5,400 seconds divided into six time intervals of 900 seconds was defined in VISSIM. The first interval represents the warm-up period, the second, third, fourth and fifth periods represent the peak hour and the last interval represents the cool down period. The vehicle inputs for each time interval are determined on the basis of processed automatic vehicle counting data obtained from the Traffic Management and Control Centre (TMCC) in Skopje. The calibration procedure described in [7] was applied for the calibration of the VISSIM model. The saturation flow was selected as the parameter for validation of the model. Because the saturation flows appear to be modeled incorrectly uniformly across the network, the parameters of the global ”driver behavior” model: average standstill distance, additive part of safety distance and multiplicative part of safety distance were adjusted to comply with the validation criteria. Obtained results with average values per approach for the selected intersection I2 in the chosen analysis period are shown in table 1 [6]. Conclusion and future work The aim of this research paper is to provide a performance evaluation of ATCS applied in today’s UTC centers. For this, a simulation framework was developed to enable a software in the loop simulation of the adaptive traffic control UTOPIA using the microscopic simulator VISSIM. It contains a new MATLAB based tool named TSCLab which can access, gather and visualize relevant data generated by VISSIM, needed for calculation of selected measures for performance evaluation of ATSC in VISSIM based microscopic simulation environment using different traffic scenarios. TSCLab provides the operators in the traffic management and control center a possibility to test the control strategy and if it is necessary, to change the algorithm parameters prior to being implemented on the field. Future work on this topic will be related to the augmentation of TSCLab to enable evaluation using a simulation model containing more controlled signalized intersections in an urban network and comparison of different control strategies. Additionally, new measures for performance evaluation especially those related to sustainable modes of transport (public transport, cycling and walking) will be considered also. ■ REFERENCES [1] I. Dakic, M. Mladenović, A. Stevanović, and M. Zlatkovic (2018), Upgrade Evaluation of Traffic Signal Assets: High-resolution Performance Measurement Framework, PROMET - Traffic&Transportation, 30(3): 323-332 [2] D. Pavleski, D. Nechoska Koltovska, and E. Ivanjko (2017), Development of TSCLab: A tool for evaluation of the effectiveness of adaptive traffic signal control systems, In: Proceedings of 5th International conference NT-2019: 386-394 [3] D. Gettman, et al. (2013), Measures of Effectiveness and Validation Guidance for Adaptive Signal Control Technologies, US Department of Transportation, Federal Highway Administration [4] C. Day, et al. (2014), Performance Measures for Traffic Signal Systems: An Outcome-Oriented Approach, Purdue University, West Lafayette, Indiana, USA [5] C. Day, D.M. Bullock, H. Li, S.M. Lavrenz, W.B. Smith, and J.R. Sturdevant (2015), Integrating Traffic Signal Performance Measures into Agency Business Processes, Purdue University, West Lafayette, Indiana, USA [6] D. Pavleski (2018), Performances Evaluation of Adaptive Traffic Signal Control in Microsimulation Environment, Master thesis, Faculty of Technical Sciences, Bitola, (in Macedonian) [7] Mayor of London (2010), Traffic modelling guidelines, TfL Traffic manager and network performance best practices (J. Smith and R. Blewitt Eds.), Transport for London Daniel Pavleski, M.Sc. Head of Unit, Department for Transport, City of Skopje, Skopje (MK) daniel.pavleski@outlook.com Performance measure Link 1 Link 2 Link 3 Link 4 Average max green time utilization ration 0.47 0.40 0.60 0.54 Average arrived vehicles per cycle 28.20 9.79 33.51 16.18 Average served vehicles per cycle 19.25 6.35 29.50 12.35 Average green occupancy ratio 0.91 0.86 1.00 0.95 Average red occupancy ratio 0.80 0.68 0.98 0.84 Average percent of arrived vehicles on green signal 18.57 11.02 22.83 10.59 Average platoon ration 0.81 0.95 0.88 0.85 Table 1: Results from TSCLab