PaperlayerView exact revision on arXiv

Source-linked AI summary

6-Layer Model for a Structured Description and Categorization of Urban Traffic and Environment

Maike Scholtes, Lukas Westhofen, Lara Ruth Turner, Katrin Lotto, Michael Schuldes, Hendrik Weber, Nicolas Wagener, Christian Neurohr, Martin Bollmann, Franziska Körtke, Johannes Hiller, Michael Hoss, Julian Bock, Lutz Eckstein

arXiv:2012.06319v2cs.OHcs.AIcs.SE

TL;DR

The paper addresses ambiguities and incomplete urban coverage in the 6-Layer Model used to structure environment descriptions for scenario-based automated-driving verification and validation. It refines and extends the model through categorized layers, guidelines, and examples, while defining an actor-independent description and identifying future adaptations for machine perception.

  • Problem

    Urban use cases exposed incomplete coverage, shifted entities and properties, changed naming, and ambiguities in the evolving 6-Layer Model, motivating a clearer definition.

  • Method

    The paper refines the 6-Layer Model with categorized items, application guidelines, explanatory examples, and an actor-independent description spanning roads, structures, modifications, dynamic objects, and environmental and digital information.

  • Results

    The refined model extends the 6-Layer Model from highway scenarios to urban environments and supports applications including verification and validation through explicit treatment of roadside structures, dynamic objects, and traffic-light states.

  • Takeaways & Limitations

    The 6-Layer Model provides a structured basis for objective environment description and subsequent scenario description without anticipating actors’ functions or behavior.

Abstract

from arXiv · show

Verification and validation of automated driving functions impose large challenges. Currently, scenario-based approaches are investigated in research and industry, aiming at a reduction of testing efforts by specifying safety relevant scenarios. To define those scenarios and operate in a complex real-world design domain, a structured description of the environment is needed. Within the PEGASUS research project, the 6-Layer Model (6LM) was introduced for the description of highway scenarios. This paper refines the 6LM and extends it to urban traffic and environment. As defined in PEGASUS, the 6LM provides the possibility to categorize the environment and, therefore, functions as a structured basis for subsequent scenario description. The model enables a structured description and categorization of the general environment, without incorporating any knowledge or anticipating any functions of actors. Beyond that, there is a variety of other applications of the 6LM, which are elaborated in this paper. The 6LM includes a description of the road network and traffic guidance objects, roadside structures, temporary modifications of the former, dynamic objects, environmental conditions and digital information. The work at hand specifies each layer by categorizing its items. Guidelines are formulated and explanatory examples are given to standardize the application of the model for an objective environment description. In contrast to previous publications, the model and its design are described in far more detail. Finally, the holistic description of the 6LM presented includes remarks on possible future work when expanding the concept to machine perception aspects.

I. INTRODUCTION

The paper motivates scenario-based validation by identifying the impracticality of distance-based testing and the need for structured environment descriptions. It traces the 6LM’s evolution from four to six layers and addresses ambiguities that arose across prior definitions.

  • Motivation: Distance-based validation of highly automated driving requires infeasible kilometers because of time and cost constraints.This motivates scenario-based verification and validation focused on safety-relevant scenarios.
  • Motivation: Scenario-based testing in open real-world contexts requires a sufficiently complete and structured description of the environment.The 6LM separates relevant environmental aspects into layers built upon each other.
  • Urban extension: Urban extension of the 6LM requires addressing roadside structures and other aspects omitted from highway-focused applications.Earlier urban categorization work was not fully incorporated into PEGASUS and subsequent publications.
  • Model refinement: Prior publications shifted entities and changed layer names, creating ambiguities that this paper seeks to resolve through a more precise 6LM definition.The paper presents the refinement as a basis for making structured environment description more accessible to researchers and safety engineers.
  • Model evolution: The model evolved from four layers to five and then six, with later work separating road-level structure, traffic infrastructure, temporary modifications, objects, environmental conditions, and digital information.The six-layer development is summarized through the cited historical comparison and related-work descriptions.

III. SCOPE AND MOTIVATION

The 6LM is intended as an actor-independent environment description that supports scenario design, ontology construction, measurement-data analysis, and comparison of scenarios. Its scope extends across settings and abstraction levels, while machine-perception description may require adaptations.

  • Scope: The 6LM focuses on categorizing the environment within the control loop involving the environment, driver or automated system, and vehicle.Its numbering structures layers rather than ranking their importance.
  • Core concepts: Entities include anything that exists, has existed, or will exist, while objects are material entities; properties assign values to entities, and relations link them.Examples include vehicle velocity, traffic-light state, and a car’s relation of driving behind another car.
  • Applications: Engineers can use the 6LM as a basis for scenario descriptions independently of whether the target language is natural, formal, or machine-readable.The model provides a characterization of the environment, entities, and properties for scenario design.
  • Applications: The 6LM can support a traffic-domain ontology by categorizing domain entities and properties into a six-layer high-level taxonomy.The paper describes this taxonomy as informally specified, with formal digital implementation left to ongoing work.
  • Applications: Test engineers use the model to record and analyze measurement data, identify influencing factors across layers, and derive scenario concepts.The environment description is intended to cover urban, rural, and highway settings at macro-, micro-, and nanoscopic levels.
  • Design boundary: The 6LM must remain unbiased and actor-independent, describing physically observable conditions without anticipating actor functions, behavior, goals, values, or norms.The paper distinguishes this environment description from situation descriptions and notes that machine-perception aspects may require adaptations.

IV. DEFINITION OF THE LAYERS

The paper defines Layer 1 as the permanent road network and traffic-guidance foundation, then explains how its geometry, markings, signs, lights, materials, and irregularities support traffic and scenario description. Properties may be assigned within layers or handled through annotations.

  • Layer framework: The layer definitions provide explanatory names and categories of entities, with an exemplary overview presented in Table 1.The overview is explicitly incomplete and is supplemented by detailed categorization and guidelines.
  • Property assignment: Properties such as position, velocity, size, material, and color can generally be described within the relevant layer, with annotations available when another placement is more convenient.The paper refers to later guidelines for resolving such placement choices.
  • Layer 1: Layer 1 describes the road network and permanent traffic-guidance objects whose properties remain unchanged within a scenario.It summarizes where and how traffic participants can drive, while non-permanent guidance objects are assigned to higher layers.
  • Layer 1: The road network includes road geometry, topology, topography, course, linkage, elevation, and lateral profile.Road markings provide lane semantics and identify areas such as shoulders, cycle paths, sidewalks, parking spaces, and keep-out areas.
  • Layer 1: Permanent traffic signs and traffic lights belong to Layer 1, while their changing states are described in Layer 6.Road markings also encode instructions such as speed limits, stopping lines, and turn arrows.
  • Layer 1 applications: Layer 1 supports microscopic mission planning through start and end points, routes, velocity profiles, and stopping requests.The same information can support macroscopic traffic-flow design and comparison of road networks.
  • Layer 1: Layer 1 includes road-surface materials, irregularities such as potholes, surface-modifying structures, and speed bumps.These additions extend the road description beyond geometry and traffic-guidance infrastructure.

B. ROADSIDE STRUCTURES (LAYER 2)

Layer 2 captures permanent roadside structures in urban environments, while Layer 3 records temporary modifications to Layers 1 and 2. Because “temporary” depends on context and may span flexible time frames, classification should be application-specific.

  • Layer 2: Layer 2 contains static roadside objects such as buildings, vegetation, walls, fences, lamps, hydrants, bollards, shelters, tunnels, bridges, and restraint systems.
  • Layer 2: Layer 2 is generally permanent at a designated position, whereas deviations or more detailed motion descriptions are assigned to Layer 3 or Layer 4.
  • Layer 3: Layer 3 represents temporary modifications of Layer 1 and Layer 2, including construction sites, temporary signs, traffic lights, markings, fallen trees, and road contamination.
  • Layer 3: Layer 3 introduces no new object classes, but instantiates existing Layer 1 or Layer 2 classes when they are non-permanent in the real world.
  • Temporal classification: “Temporary” cannot be assigned a fixed threshold because regulations provide only contextual indications and modifications may persist across flexible time frames.
  • Layer separation: Layer 1 and Layer 2 describe spatial and non-temporal properties, while Layer 3 captures construction-related changes that modify them.

D. DYNAMIC OBJECTS (LAYER 4)

Layer 4 is the first time-dependent layer and describes dynamic objects whose movements or other state changes evolve within a scenario. Its urban scope includes both traffic participants and stationary objects capable of movement.

  • Layer 4 definition: Layer 4 is the first layer with a time-dependent description and includes movable objects whose movements evolve through trajectories or maneuvers.
  • Dynamic objects: Dynamic objects include entities that move or can potentially move, including parked vehicles, standing pedestrians, and garbage cans awaiting pickup.
  • Dynamic objects: Layer 4 was renamed “Dynamic Objects” because it also covers time-dependent state changes not necessarily associated with movement, such as changing road-marking visibility.
  • Urban examples: Urban Layer 4 examples include vehicles, motorcycles, bicycles, pedestrians, trams, animals, garbage cans, balls, and falling trees.
  • Scope boundary: The 6LM excludes non-physical interactions, goals, and values because it describes observable physical environment elements.

E. ENVIRONMENTAL CONDITIONS (LAYER 5)

Layer 5 describes globally perceptible environmental conditions, including weather, atmospheric, lighting, and road-weather states. Effects that become dynamic or persist as constant modifications are assigned to Layers 4 or 3, respectively.

  • Layer 5 definition: Layer 5 contains weather, atmospheric, lighting, and road-weather conditions such as precipitation, fog, wind, cloudiness, wetness, ice, and sun position.
  • Perceptibility: Layer 5 is limited to globally perceptible conditions, while actor-dependent occlusions can be derived for individual actors from the 6LM information.
  • Effects across layers: Environmental effects that induce movement or other dynamic changes are described in Layer 4, such as wind-driven cones, moving leaves, or progressively obscured markings.
  • Effects across layers: A road marking covered by snow throughout the scenario is described in Layer 3 because the modification is constant and lacks a dynamic component.

F. DIGITAL INFORMATION LAYER 6

Layer 6 describes information exchange, communication, and cooperation based on digital data. It includes V2X messages, signal coverage, and digitally encoded traffic-management states, while entity classification remains independent of information source.

  • Layer 6 definition: Layer 6 focuses on digital information exchange, communication, and cooperation between vehicles, infrastructure, or both.
  • V2X information: Layer 6 includes V2X information such as road closures, extreme-weather warnings, truck-to-truck platooning messages, and wireless signal coverage and strength.
  • Information source: The same vehicle braking action and traffic jam remain Layer 4 entities regardless of whether the information is observed by sensors or received through V2X.
  • V2V example: A V2V message can supplement an occluded intersection scene by communicating a vehicle’s intention to turn left when that intention is otherwise unavailable.
  • Traffic management: Traffic-management states, including traffic-light and switchable-sign states, are assigned to Layer 6 because they are encoded as digital information.

V. GUIDELINES FOR THE 6-LAYER MODEL

The guidelines organize the 6-Layer Model by separating spatial and temporal description, locating temporary changes and state changes appropriately, and assigning properties consistently across layers.

  • Layers 1–3 describe spatial aspects without time-variable content, while time-based descriptions begin at Layer 4.
  • Layer 3 contains temporary changes to Layers 1 and 2 that remain fixed throughout the scenario.
  • State changes begin at Layer 3, and from Layer 4 upward they may also be time-dependent.
  • Entities with potentially time-dependent properties belong on Layer 4 or higher, although some of their properties may remain constant.
  • An entity’s properties may occupy different layers, but each specific property should remain on one layer and be placed where it best matches the layer description.

B. EXPLANATORY DESCRIPTION AND EXAMPLES FOR EACH GUIDELINE

The explanatory examples show how the guidelines support reusable spatial descriptions, consistent treatment of properties and states, actor-independent categorization, and explicit cross-layer influences.

  • Separating spatial Layers 1–3 from temporal Layers 4 and above allows fixed road and infrastructure descriptions to be reused across recordings.Layer 3 records modifications to otherwise constant Layers 1 and 2.
  • State changes in Layers 1 and 2 are assigned to Layer 3 when the modification is visible but remains fixed for the scenario.
  • Movable entities are placed from Layer 4 upward because their positions can change, while constant attributes such as size and color may also be retained.
  • Traffic lights span layers by having spatial placement in Layer 1 and time-varying switching states in Layer 6.The example distinguishes an entity’s location from its changing control state.
  • Annotations can record supplementary roles or simplify detail, such as noting an officer’s regulatory duty or a bush moving in wind without modeling its motion fully.
  • The model assigns properties according to their relevant influences, illustrated by friction depending on road, condition, and contacting material rather than one global layer.
  • Because the 6LM is actor-independent, it describes the environment without encoding individual occlusions, which can be calculated later for selected participants.
  • Layer numbering structures categories rather than ranking importance, and influences may run between earlier and later layers.Examples include dense traffic affecting shoulder use and weather affecting road-marking visibility.

VI. EVALUATION THROUGH REAL-WORLD DATA

A drone-recorded urban intersection demonstrates the 6LM’s application across spatial, temporal, environmental, and digital layers, while also exposing which information is difficult to recover from recordings.

  • The evaluation applies the 6LM to a drone recording of a four-armed urban intersection, using a sequence of frames to represent temporal scenario development.
  • Layers 1 and 2 capture road geometry, markings, traffic signs, road-surface irregularities, street lamps, buildings, roadside structures, bicycle stands, and vegetation.
  • The recording contains no Layer 3 content because no temporary modifications to Layer 1 or Layer 2 elements are present.
  • The invariant Layer 1 and Layer 2 description can be retained when additional recordings are made at the same intersection.
  • Layer 4 includes moving and stationary vehicles, bicycles, and pedestrians, including pedestrians using the crosswalk.
  • Layer 5 can identify dry road conditions and visible shadows, but a complete environmental description is difficult without additional information.
  • The intersection has no traffic lights, switchable traffic signs, V2X infrastructure, or known digitally connected participants, leaving no Layer 6 information to depict.

VII. FUTURE WORK

Future work targets a domain ontology for standardized 6LM descriptions and examines adaptations needed for machine-perception-oriented environment modeling.

  • The authors plan to implement the 6LM in a domain ontology for verification and validation of highly automated vehicles.
  • The ontology will formalize traffic entities, properties, and relations, including subclasses for traffic participants and specific weather conditions.
  • Machine-perception descriptions may require adaptations because perception aspects cannot necessarily be represented actor-independently.
  • Perception-oriented modeling may need more detailed descriptions of surrounding materials, reflections, and contamination than driving-function descriptions require.
  • Additional work will investigate data sources for enriching recorded test data.

VIII. SUMMARY

The paper refines the 6-Layer Model for environment description and extends it beyond highway scenarios to urban traffic. Its definitions and guidelines support standardized, objective descriptions that can underpin scenario descriptions and ontologies.

  • The refined 6-Layer Model extends environment description from highway scenarios to more complex urban applications.Urban use required concepts for roadside structures, additional dynamic objects, and traffic-light states.
  • Definitions and guidelines establish a standardized, generally usable, unbiased, and objective categorization of the environment.The work covers all layers and provides explanatory examples to support application of the model.
  • The 6-Layer Model provides a structured basis for scenario descriptions and ontologies.
Loading 2012.06319v2…