SCIENCE & RESEARCH Automation International Transportation | Collection 2020 58 Implementation of autonomous vehicle onto roadways A step to a Theory of Automated Road Traffic Road map of progress, Automation chain, Scenery finding, Scenario construction, Movement spaces, Interaction scenes At present, automation researchers and automotive component developers perceive the car to be a solitary object that constitutes a sort of singularity, which both triggers and copes with events onto roadways. As far as we know, the setting in which events occur along a road and require automated responses has so far been studied only at a highly abstract level and only for singular events that occur directly in the course of traffic. No comprehensive analysis has so far been attempted that discusses structures of the physical setting in greater detail both objectively and in terms of spatiality and that looks into their disposition for anthropogeneous intervention in response to autonomous vehicle movement. Heinz Doerr, Andreas Romstorfer C urrent tenders on the subject follow a road map proposed by the technology development community triggered by automobile OEM. The goal is to have automation technologies for sensors, the software that interprets their detections and the responding electronic controls prove their suitability and safety for a dynamic vehicle in actual operation on selected test runs (like special testing grounds or on sections of public roads permitted by the authorities). During this early phase of developing automations technology, it is sensible to choose test settings that match the components state of development, especially since government authorities need to issue special permits for this purpose. The road map of technologies for automated driving defines five steps (SAE levels) until full autonomy, i.e. ultimately driverless driving. The higher the degree of automation in a car the greater will be the need and necessity to exchange data with other vehicles (v2v) in the immediate surrounding as well as with those in the setting through which the car passes and from which cars enter the main traffic flow from stationary traffic facilities. Similarly important is the need to have cars linked in an overriding network of traffic flows, e. g. in order to prevent congestions, all of which requires an exchange of data with-the traffic infrastructure (v2i resp. i2v) and its timely capacity management (v2tm2v). Spatial spheres of implementation and evaluation The next step is to break down the “setting” by its functional properties and structural features and to identify its in-depth effect on the street. What is the application perspective of automatization for all traffic participants, if they are motorized or nonmotorized, riding by public transport or individual mobile without using a car. Such a setting is not concentrated on a single vehicle and on driver’s comfort only, but it covers the mobility needs of all traffic participants and takes care of their interests in critical interactions between traffic participants on a roadway. Hereby reference is made to affected spaces, which postulates spatial categories in respect of perimeters and scale as follows: Surrounding onto roadway (as “Playing ground of scenes”) That concerns traffic spaces onto the roadway used by autonomous and traditional cars as well as by other authorized traffic participants. Traffic flows are regulated by planning and management measures and the local regulations are signalized to the traffic participating groups. Setting of Scenery (as traffic generating and otherwise intervening land use structures) Hereby the public street and the land use of its adjoining spaces are essential, i. e.- urban spaces that have the potential to- generate traffic as well as a risk potential- for incidents to impact on the traffic flow. Environment Influences resulting from the perimeter, whether near or far but not clearly delimitable, along a traffic corridor that may elude any standardised predictions (e.g. unreliable pedestrian flows, plumes and fumes, flocks of birds, changing lighting conditions etc.). Methodical approaches for generating test scenarios Such a spatial differentiation, which maps both interaction spaces and affected mobility areas, may serve as a practicable start- based on multiple methodological approaches, which allow evaluations and interpretations derivable from several data sources, including: • image analysis of traffic scenes (e.g. evaluations of webcams used in traffic monitoring), • spatial analysis of traffic generating sceneries, which includes mobility patterns of the inhabitants and the working people there, by way of geographic information systems GIS, • local monitoring of traffic relevant interfaces within the road network (e.g. at neuralgic locations that tend to be overlooked by stationary monitors, such as an exit from an underground car park or an intersection of main roads), • traffic detection at counting points (e.g. evaluations to identify traffic intensity across Automation SCIENCE & RESEARCH International Transportation | Collection 2020 59 time and vehicle mix by common vehicle categories), • analysis of road accidents, which provides pointers at risks that might be prevented or alleviated, but possibly even could be- enhanced by autonomous driving actions. Generating a scenario means to implement future aspects of driving manœuvres due to automation function of vehicles in scenes of daily traffic actions as they can be observed nowadays. According to the physical and traffic organisational preconditions presented by sceneries along a roadway standard event scenarios when interacting traffic participants behave regularly to the rules and eventuality scenarios when “unfortunate circumstances” occur due to irregular behaviour of players or unanticipated interventions from outside will be generated. They may result in some sort of classification of critical respective risky potentials along an evaluated route in the road network. Finding of sceneries for designing test settings To this end, streets as examples are used that run radially from the highly urbanized spaces via suburban structures to the fringe of the agglomerations because they contribute a large variety of realistic frame conditions in terms of roadway’s topography and daily traffic events. Therefrom implementation conditions can be identified for the use of automation functionalities in road traffic. Ultimately this is to yield testing tools such as a catalogue of testing criteria for licensing automation functionalities in vehicle types when the driver’s responsibility will be delegated in whole or in part to the automation system depending on the automation-level (SAE) of the car. Besides finding of revealing sceneries as examples for test settings one can discover road sections along the route, which are suitable and adaptable for automatization of road traffic. On the one hand, sections with four lanes on separated carriageways or tunnel sections of expressways are predestined for that as well as motorways in general. With them external guiding systems to establish interconnectivity between infrastructure and vehicles could easily installed, if the operator of the road network would bear the investment costs (figure 1). On the other hand, urban main roads are characterised by a multitude of surface marks and traffic signs at the roadside for regulating traffic flow as shown in figure 1 and figure 2. These local regulations should been recognised “visually” by the detection functionality or could be received as signals send by the stationary signs equipped with transponders. Besides of that such information could feed in an on-board database of the vehicle, which has to be the latest. System expansion to non-automated and non-interconnected traffic participants Which traffic participating groups will be concerned when vehicles will be fully or partly equipped with automated functionalities allowing hands-off or even eyes-off to the driver? To whose responsibilities regulations have to be addressed if malfunctions of the technologies occure and cause insecurity or incidents. All those crucial questions evoke need for clarification, before automation conquers the mixed used public traffic spaces. The realization of personal mobility must be perceived as a vested right of all mobility groups considering their different needs and opportunities. Issues of how to design future mobility touch not just on technological aspects such as digitalising cars and making them autonomous, but also need to be linked to the contrarian idea of combining central control over traffic with individual freedom of movement within the traffic system. Fitting out means of transport, and in particular passenger cars, with ever more digital equipment changes the conditions for exercising the right to mobility for all groups. 2 Figure 1: A circular driving survey to find sceneries for testing arrangements within the Northeast of Vienna Methodical approaches for generating test scenarios Such a spatial differentiation, which maps both interaction spaces and affected mobility areas, may serve as a practicable start based on multiple methodological approaches, which allow evaluations and interpretations derivable from several data sources, including:  image analysis of traffic scenes (e.g. evaluations of webcams used in traffic monitoring), High-density urban precinct (Floridsdorf) P e r i u r b a n f r i n g e Suburban miscellaneous landuse structures Figure 1: A circular driving survey to find sceneries for testing arrangements within the Northeast of Vienna All figures by the authors SCIENCE & RESEARCH Automation International Transportation | Collection 2020 60 Looking at Interaction Spaces as playing ground of mobility In a next step, traffic automation technology will expand to encompass non-automated motorists, such as drivers of older vehicles that have not been retrofitted, as well as non-motorized traffic participants who couse, pass along or cross traffic lanes. These groups of traffic participants could be upgraded in the use of ICT and incorporated in the traffic network, e.g. via their smart phones, but this involves issues that go far beyond the technological discourse and touch principal societal aspects. Scenarios could have been derived from daily observations of traffic scenes by coincidentally interacting traffic participants with respect to an assumed mixture of vehicles on different levels of automatization and powertrain performance (scene in figure 2). For this purposes overlookable interaction’s spaces have to be delimited that could be defined by physical and traffic guiding conditions of the roadway using geoinformation data in high solution and the graph theory as attribution order of roads. In such a way test objects and their playing grounds can be fixed to design testing arrangements. One has to distinguish between tasks of autonomizing cars like distance keeping or collision avoidance and the technological means to fulfill them. Automotive engineers prefer to talk about technologies and their deployment but less about the real world implementation of autonomized vehicles in their complex relationships to other mobility groups. Focussing back to the vehicle’s automation system, the functionalities form a chain of decision-preparing resp. decision-making procedures enabling the car to steer movements autonomously, which has to be done in splits of seconds permanently. That goes beyond of assistance-systems at what driver’s responsibility fully remains in the case of incidences. Moreover, it requires an autonomous so-called “neuronal” networking of functionalities based on deep learning which has gained artificial intelligence by while. A holistic view on the issues of-automated road traffic As a key to analyse scenes and predict interactions between traffic participants onto the roadway it will be necessary to investigate the complexity and the impact of frame conditions. Moreover is a need to create a terminology derived from daily traffic actions beyond an automotive technological speech. Beginning with the detection of the disposable trajectory field and the ranging of moving objects, which could intervene the driving of the more or less autonomized car, the automation chain of the vehicle has to predict the moving spaces of the opponent traffic participants around. In this context trajectories of them are not only likely ideal lines marked out but they form also a hardly to calculate moving space of mobile objects. At this end a moving space in 3-D as defined here contains the potential of movement actions onto a (normally by the traffic rules designated) “playing-ground”. Categorization of subjects as required for scenario construction Making scenario construction operable there is a need of systematic categorization of relevant subjects, of which characteristics and attributes more or less are suited to master challenging tasks of automatized driving. As they are (maybe in an incomplete listing): • Vehicle’s automation equipment corresponding to the SAE-levels • Vehicle’s motorization corresponding to vehicle utility classes and automotive brands • Roadway’s tracing characteristics embedded in the landscape 4 Figure 2: Sailing Interaction Space of a coincidental group of different automated vehicles as scenario draft Vehicle-Roadway-Relations V2I (as frame conditions by the road infrastructure affecting the driving actors) Passenger car ( a2 ) of a vehicle class (B 3 ) moving onto roadways of a distinctive category (III 2(G) ): B 3a2 _III 2(G) Highly automated car Kfz ( a3 ) of a vehicle class (B 3 ) moving onto a roadway of category (III 2G ): B 3a3 _III 2G Partly autonomous moving car ( a4 ) of a light vehicle class (B 1 ) onto a special equipped roadway (III 2G ): B 1a4 _III 2G Interactions between moving vehicles V2V (bilateral and multilateral setting of conditions) Initial and reactive actions by involved vehicles of different vehicle classes equipped on different levels of automation standards within the current sailing interaction space B 1a4 -B 3a2 Traffic actions by the involved vehicles (coincidental conditions of traffic behaviour by different automated cars) Autonomous trajectory seeking: Prediction to beware counter moving traffic: Overtaking manœuvre without driver´s assistance tool: Side distance keeping in respect of counter moving traffic: Side distance keeping in respect of parallel moving traffic: Forward braking distance control: Forward collision alert: Backward distance alert in respect of following traffic: Driver´s awareness, presence or absence (human role of driver´s tasks controlling a vehicle) Driver fully controlling vehicle movement: Driver highly supported by Driving Assistance Systems: Driver off duty (hands and eyes off) but ready for take over (Human-Machine-Interaction): Not figured in that scenario: No driver present: Vehicle carries passengers: No Person in the car: Infrastructure Equipment for Interconnectivity I2V Guiding wire line surface integrated: Guiding wire line kerb integrated: Wireless option: transponder installed on lighting pylons: 5G-Mobile Radio Net: Sailing interaction space of a coincidental group of vehicles 5 G B 1a4 -B 3a2 B 1a4 _III 2G D a2 -B 1a2 B 1a2 _III 2G D a2 _III 2G B 2a2 _III 2G C a3 _III 2G _III 2G B 3a2 _III 2(G) B 3a3 _III 2G Basic scene(ry) … … Figure 2: Sailing Interaction Space of a coincidental group of different automated vehicles as scenario draft Scenery: Hamburg-Diebsteich, a scene caught on a Monday’s noon in June Automation SCIENCE & RESEARCH International Transportation | Collection 2020 61 • Roadway infrastructure to master typical traffic function within the road network • Roadway surrounding tract in respect of traffic generating land uses and external risks • Rhythm of daily traffic flows occurring onto sections of the road network • Local caused circumstances by the natural environment and the built-up area • List of mobile groups of traffic participants and their movement behaviours and handicaps • List of vulnerable road users and their special safety needs • List of supporting means of moving which are not or weakly motorized • Standardized scenes of driving manœuvres as they can be often observed • Systematization of interactions between traffic participants Approaches to generate traffic oriented scenarios First of all assumptions have to be declared about the expected future traffic scenes, what has been derived from observations of daily traffic scenes and what is anticipated as future technological options of automated car driving: • The estimated deployment of automation products as car-equipment on the sales-market as hint at the mix of vehicles on different levels of automation for a certain horizon of implementation (until 2030, 2040 and so on) • Selecting tracts of land-use-environments with typical traffic occurences onto the road network • In this context, future applications of automation technologies and functions integrated in vehicles should be interrogated in respect of their applicability for a tract like a suburban residential area or an overcrowded central district. • Finding of representative sceneries within them • Cutting interaction boxes of road-sections within such sceneries (like the surrounding of an urban intersection as figured out) from the view-point of daily road traffic • In a further step one can use the tool-box of listings as mentioned above to generate variants of scenes in which traffic participants could be periodically or sporadically or potentially involved in risky traffic interactions • Traffic scenes have a volatile constituency of participants involved and a variable shape of action area. It deals with “Sailing Interaction Spaces” in respect of processing and timing bilateral and multilateral actions. That reminds one of an amoeba. Construction of an In-situ- Scenario as guideline for test arrangements Scenario generating means combining static and dynamic recurring frame conditions as rules of an interaction box respectively. “playing ground” with the behaviour of traffic participating players coping with tasks of driving or moving in a- consecution of interactions between them. The scenario depicted in figure 4 dealt with the task how an automated driving system would react if vulnerable road users will be the initial actors on the scene. The short story hereby tells a scenario of interactions between a cyclist as initiative actor at the beginning, a heavy-duty truck as reactive actor and pedestrians as independent players. The interactions between them take place in an interaction box as- part of an intersection as depicted in figure-3. The traffic light gives the starting signal when it shows green release for going straight ahead or turning to the right. The initial actor is a cyclist waiting for green light at the stop line while a heavy truck is approaching. The cyclist has two options to direct himself: forwards straight ahead or turning right. If both participants are going for turning towards right, it will be a tremendous challenge for an autonomized heavy vehicle detecting the forefield, predicting the behaviours of the others and controlling his own driving dynamics. This story leads to a setting for test arrangements proving the automatic system of the vehicle based on requirements of the “real world” which cannot be done convincingly by computer-simulation only. 6 Tramway Stop Residential block Direction downtown Direction outskirts Car dealer Commercial parking lot Furniture sales Direction urban transit motorway 6 Tramway Stop Residential block Direction downtown Direction outskirts Car dealer Commercial parking lot Furniture sales Direction urban transit motorway Figure 3: Mobility groups meet onto an intersection of urban main roads frequenting their moving spaces regularly while traffic light shows green for northbound-southbound passage and for turning right of cars crossing pedestrians pathways. SCIENCE & RESEARCH Automation International Transportation | Collection 2020 62 Conclusion and outlook Therefore transferring information about impacts on traffic behaviour and side effects on mobility activities will be essential. Better information given to the involved groups would not be sufficient if thereby a lack of basic knowledge about traffic actions and events seems to be manifest. Before enlightning information can given it requires a wide range of testing arrangements based on a systematically developped theory of automated road traffic. In the end, not only automotive engineers and suppliers should be entrusted with automatization. Several disciplines and expert groups as well as representatives of somehow involved mobility groups should be participated in the procedures of implementation. That refers nation-wide to the general regulatory frame, which has to put into force or takes aim at regional or local regulations to be edicted by public administrative entities. Somewhere the rebuilding of road equipment could be necessary to provide interconnectivity between cars and infrastructure. Other where the traffic network has to be reorganised in order to minimize conflicts between the diversified groups of traffic participants resp. road users. So one can conclude that advanced automatization of vehicles and other transport means will affect not only car owners and users, it will provoke a system change in usage the public roadways and the procedures of road traffic as well. ■ SOURCES Heinz Dörr, Viktoria Marsch, Andreas Romstorfer (2017): Automatisiertes Fahren im Mobilitätssystem. Ein Spannungsbogen zwischen Ethik, Mobilitätsausübung, technischem Fortschritt und Markterwartungen. Internationales Verkehrswesen (69) 3, pp. 40-44 Heinz Dörr, Viktoria Marsch, Andreas Romstorfer (2017): Automatisiertes Fahren in urbaner Umgebung. Herausforderungen für die Stadt- und Verkehrsplanung. Transforming Cities (2) 3, pp. 47-53 Heinz Dörr, Viktoria Marsch, Andreas Romstorfer (2018): Automatisierter Straßenverkehr und spurgebundener ÖPNV. Betroffenheiten, Verantwortlichkeiten, Handlungsbedarfe. Der Nahverkehr (36) 3, pp. 58-65 Heinz Dörr, Viktoria Marsch und Andreas Romstorfer (2019): Automatisiert bewegen durch Stadt und Land - Gesellschaftliche Implikationen der Implementierung von ITS-Technologien in das Verkehrsgeschehen des zukünftigen Mobilitätssystems.. REAL CORP 2019 24th International Conference on Urban Planning and Regional Development in the Information Society (Karlsruhe Institute of Technology), Tagungsband pp.111-121, online: www.corp.at Heinz Dörr und Andreas Romstorfer (2020): Theoretische und praktische Ansätze zur Implementierung des automatisierten Straßenverkehrs in das Mobilitätssystem. In: Heike Proff (Hrsg.): Neue Dimensionen der Mobilität. Tagungsband zum 11. Wissenschaftsforum Mobilität 2019 der Universität Duisburg-Essen, pp. 719-743 Heinz Dörr, Dr. Consulting engineer spatial and traffic planning, arp-planning. consulting.research, Vienna (AT) heinz.doerr@arp.co.at Andreas Romstorfer, Dipl.-Ing. (FH), MA arp-planning.consulting.research, Vienna (AT) a.romstorfer@arp.co.at Figure 4: By vulnerable road users initially deployed scenario of detection and prediction in the phase of green-release for straight ahead or turning right traffic of heavy-duty vehicles, cyclists and pedestrians Consecution of scenes t0 to t5 by interacting actors: t5: Pedestrian´s crossing pathway cleared by pedestrians, truck can pass t4: Pedestrians initial, truck reactive, cyclist waiting for passing to bicycle stripe t3: Cyclist forced to be waiting - following truck reactive stopping t2: Pedestrians begin with crossing - cyclist reactive for a while stopping t1: Cyclist starts selecting direction - truck reactive t0: Traffic light switches to green Type of Vulnerable Road User Group: Cyclist waiting at the stop line Pedestrian unrestrained but hidden Pedestrian waiting but in risky position Person with child running after Person with child pushing a perambulator Person moving with a wheelchair Presumed trajectories of them Safety distance space required Detection and prediction requirements: Proposed trajectory detected by truck Vehicle classes and automation levels: Proposed trajectory observed by cyclist Trajectory anticipation of subsequent scene By an automated truck By human awareness Bus highly assisted a3: Distance keeping in focus Lane keeping and clear road surveillance Grey marked cars and cycles not directly involved Cars are sorting for proposed direction Side and backward distance control Function of Interaction-box: Side surveillance to prevent unlucky incidents Part of an intersection for inflowing traffic Fan out and sorting directions of traffic Entry into interaction-box a2 a3 a2 a2 Static Interaction-box of southern inflow section with sailing interaction space a4 End of interaction-box Tramway stop platform Pedestrian´s walk Truck with trailer highly automated a4 a3