In this work, we propose a novel taxonomy to partition the problem of testing advanced driver assistance systems (ADAS) into three basic dimensions. These dimensions are detailed and confirmed with recent research. Our framework permits the consideration of open research questions which have to be answered to pave the way for future highly automated driving. Despite the importance of this problem, a similarly comprehensive and structured survey has to the best of the authors' knowledge not been developed before.

/pdf/testing-of-advanced-driver-assistance-towards-automated-1b10uxt2hh.pdf

Testing of Advanced Driver Assistance Towards Automated Driving: A Survey and Taxonomy on Existing Approaches and Open Questions

This paper capitalizes on the increasingly high relevance gained by data-intensive technologies in the development of intelligent transportation system, which calls for the progressive adoption of adaptive, self-learning methods for solving modeling, simulation, and optimization problems. In this regard, certain mechanisms and processes observed in nature, including the animal brain, have proved themselves to excel not only in terms of efficiently capturing time-evolving stimuli, but also at undertaking complex tasks by virtue of mechanisms that can be extrapolated to computer algorithms and methods. This paper comprehensively reviews the state-of-the-art around the application of bioinspired methods to the challenges arising in the broad field of intelligent transportation system (ITS). This systematic survey is complemented by an initiatory taxonomic introduction to bioinspired computational intelligence, along with the basics of its constituent techniques. A focus is placed on which research niches are still unexplored by the community in different ITS subareas. The open issues and research directions for the practical implementation of ITS endowed with bioinspired computational intelligence are also discussed in detail.

Bioinspired Computational Intelligence and Transportation Systems: A Long Road Ahead

The validation and verification of cognitive skills of highly automated vehicles is an important milestone for legal and public acceptance of advanced driver assistance systems (ADAS). In this paper, we present an innovative data-driven method in order to create critical traffic situations from recorded sensor data. This concept is completely contrary to previous approaches using parametrizable simulation models. We demonstrate our concept at the example of parametrizing lane change maneuvers: Firstly, the road layout is automatically derived from observed vehicle trajectories. The road layout is then used in order to detect vehicle maneuvers, which is shown exemplarily on lane change maneuvers. Then, the maneuvers are parametrized using data operators in order to create critical traffic scenarios. Finally, we demonstrate our concept using LIDAR-captured traffic situations on urban and highway scenes, creating critical scenarios out of safely recorded data.

Data-driven simulation and parametrization of traffic scenarios for the development of advanced driver assistance systems

Accurate probabilistic prediction for intention and motion of road users is a key prerequisite to achieve safe and high-quality decision-making and motion planning for autonomous driving. Typically, the performance of probabilistic predictions was only evaluated by learning metrics for approximation to the motion distribution in the dataset. However, as a module supporting decision and planning, probabilistic prediction should also be evaluated from decision and planning perspective. Moreover, the evaluation of probabilistic prediction highly relies on the problem formulation variation and motion representation simplification, which lacks a formal foundation in a comprehensive framework. To address such concerns, we provide a systematic and unified framework for the analysis of three under-explored aspects of probabilistic prediction: problem formulation, representation simplification and evaluation metric. More importantly, we address the omitted but crucial problems in the three aspects from decision and planning perspective. In addition to a review of learning metrics, metrics to be considered from planning perspective are highlighted, such as planning consequence of inaccurate and erroneous prediction, as well as violations of predicted motions to planning constraints. We address practical formulation variations of prediction problems, such as decision-maker view and blind view for viewpoint, as well as reactive prediction for interaction, so that decision and planning can be facilitated.

/pdf/probabilistic-prediction-from-planning-perspective-problem-10tws6d9e1.pdf

Probabilistic Prediction from Planning Perspective: Problem Formulation, Representation Simplification and Evaluation Metric

A highly or fully automated vehicle (SAE J3016 Taxonomy [1]) will be a new technology when it enters public traffic. Nevertheless the safety of this new technology will be compared to that of human drivers assisted by advanced driver assistant systems (ADAS [2]). This combination of high safety requirements and new, higher levels of complexity leads to the so called “approval trap” that could befall the introduction of automated vehicles. The “approval trap” describes the situation in which there is no known way to prove that an automated system could match a human driver in terms of safety, even if such a system should exist. It is claimed that the state of the art in vehicle testing needs to be enhanced to bring these vehicles to everyday use [3]. Regarding these challenges this paper motivates a new runtime validation method called Virtual Assessment of Automation in Field Operation (VAAFO). Based on that, the VAAFO concept is discussed with the aid of a developed architecture and use cases that illustrate the principles. The benefits and drawbacks are highlighted.

Virtual Assessment of Automation in Field Operation – A New Runtime Validation Method

This paper presents the use of advanced perception systems for the automated generation of simulated driving scenarios. We describe our advances in the fields of laser scanning processing for reference generation, and illustrate the use of reference data for constructing simulated virtual scenarios that can be loaded, manipulated and used within a commercial simulator.

Automatic scenario generation based on laserscanner reference data and advanced offline processing

Intelligent vehicle systems such as Advanced Driver Assistance Systems (ADAS), Intelligent Transportation Systems (ITS) and automated driving systems consist of increasing numbers of sensor technologies as well as increasingly advanced algorithms for sensor fusion, object tracking, object classification, risk estimation, driver status recognition and vehicle control. It is rapidly becoming infeasible to check the performance of each new (sensor) system in the traditional way: By performing test drives, storing data, manually labelling the data for reference, and manually evaluating the results. One of the approaches to address these difficulties is to install a reference sensor system on the test vehicle in addition to the prototype sensor system (device under test). The recorded data from the reference sensor system are processed – partly or fully-automatically to create reference scenarios, based on automatic object labelling and automatic event identification. Based on these reference data, the device under test can be automatically and objectively evaluated. The reference data from the reference sensor system can now be converted into a set of virtual scenarios which can be used within a CAE environment. These simulated “ground-truth reference scenarios” offer a platform for engineers to quickly check the consequences of design changes to the device under test, and allow engineers to subject the device under test to a wide variation of virtual traffic scenarios.

Automated Generation of Virtual Driving Scenarios from Test Drive Data

An evaluation of sensor and function performance of an automated driving system as well as behavior analysis in certain situations are important steps on the road to automated driving. The prerequisite is the filtering of relevant situations with an automated and reliable offline classification of maneuvers in large amounts of driving environment data. In this paper, a dual approach for detecting lane changes in laser scanner environment data is presented and evaluated. It uses two classifiers combined, a probabilistic and one based on fuzzy logic, leading to a higher confidence of the results with lower false positive detections. In addition, different probabilistic methods for the former classifier are compared and evaluated.

Dual approach for maneuver classification in vehicle environment data

This paper presents the development of a Traffic Collision Avoidance System (TCAS) for trams based on state-of-the-art automotive laser scanning technology. Our work investigates TCAS regarding required functionalities and current efforts, and proposes an innovative solution for trams that is based on Ibeo laserscanners and digital maps. The different components of the proposed system are presented together with a description of upcoming field tests to be carried out in several European countries.

Towards the development of a laserscanner-based collision avoidance system for trams

This paper presents the follow-up of the Traffic Collision Avoidance System (TCAS) for trams based on state-of-the-art automotive laser scanning technology introduced in [10]. This work investigates TCAS regarding required functionalities and current efforts, and proposes a solution for trams that is based on Ibeo laserscanners and digital maps. The different components of the proposed system are presented together with a description of preliminary results during field tests.

Roman Katz

Papers

Automatic scenario generation based on laserscanner reference data and advanced offline processing

Automated Generation of Virtual Driving Scenarios from Test Drive Data

Dual approach for maneuver classification in vehicle environment data

Towards the development of a laserscanner-based collision avoidance system for trams

Collision Avoidance System for Trams using Laserscanners