Abstract

International audience; The modes of reasoning which are used in the context of safety analysis and the very nature of knowledge about safety mean that a conventional computing solution is unsuitable and the utilization of artificial intelligence techniques would seem to be more appropriate. Our research has involved three specific aspects of artificial intelligence: knowledge acquisition, machine learning and knowledge based systems (KBS). Development of the knowledge base in a KBS requires the use of knowledge acquisition techniques in order to collect, structure and formalizes knowledge. It has not been possible with knowledge acquisition to extract effectively some types of expert knowledge. Therefore, the use of knowledge acquisition in combination with machine learning appears to be a very promising solution. This paper presents the result of these two research activities which are involved in the methodology of safety analysis of guided rail transport systems.

Abstract

International audience; Air Traffic Management involves many different services such as Airspace Management, Air Traffic Flow Management and Air Traffic Control. Many optimization problems arise from these topics and they generally involve different kinds of variables, constraints, uncertainties. Metaheuristics are often good candidates to solve these problems. The book models various complex Air Traffic Management problems such as airport taxiing, departure slot allocation, en route conflict resolution, airspace and route design. The authors detail the operational context and state of art for each problem. They introduce different approaches using metaheuristics to solve these problems and when possible, compare their performances to existing approache

Abstract

The two first CEAS (Council of European Aerospace Societies) Specialist Conferences on Guidance, Navigation and Control (CEAS EuroGNC) were held in Munich, Germany in 2011 and in Delft, The Netherlands in 2013. ONERA The French Aerospace Lab, ISAE (Institut Superieur de l’Aeronautique et de l’Espace) and ENAC (Ecole Nationale de l’Aviation Civile) accepted the challenge of jointly organizing the 3rd edition. The conference aims at promoting new advances in aerospace GNC theory and technologies for enhancing safety, survivability, efficiency, performance, autonomy and intelligence of aerospace systems. It represents a unique forum for communication and information exchange between specialists in the fields of GNC systems design and operation, including air traffic management. This book contains the forty best papers and gives an interesting snapshot of the latest advances over the following topics: l Control theory, analysis, and design l Novel navigation, estimation, and tracking methods l Aircraft, spacecraft, missile and UAV guidance, navigation, and control l Flight testing and experimental results l Intelligent control in aerospace applications l Aerospace robotics and unmanned/autonomous systems l Sensor systems for guidance, navigation and control l Guidance, navigation, and control concepts in air traffic control systems For the 3rd CEAS Specialist Conference on Guidance, Navigation and Control the International Program Committee conducted a formal review process. Each paper was reviewed in compliance with standard journal practice by at least two independent and anonymous reviewers. The papers published in this book were selected from the conference proceedings based on the results and recommendations from the reviewers. Document type: Part of book or chapter of book

Abstract

This document does not aim to review automated vehicles' developments and technologies but automated functions in existing vehicles. Therefore this document reports the results from the benchmarking search carried out within A1.1 for systems that are relevant to the project's preliminary Use Cases of ADAS&ME project. This report reflects the work carried out in order to reach MS2 (M3) of ADAS&ME and the literature review carried out for each affective state. However, it acted as a living document until the submission of D1.1 (M6) to be incorporated in the main body of D1.1. In general, benchmarking is considered a continuous and systematic approach to measure the processes and outputs against the most relevant (and occasionally) toughest competitors in the business, or at least the most prominent/ profitable ones. This is a so called competitive benchmarking approach. Three different types of results are included in this document: a) benchmarking of driver and rider monitoring systems, b) literature search and reviews of literature on drive/rider affective states as defined within DoA and UCs, and, lastly, c) alignment of the SoA collected, systems benchmarked; with key selected organization's priorities and roadmaps about automated vehicle systems and functions. The deliverable opens with the introduction to automation, the objectives of the document, the working group that was involved in A1.1 and worked towards this compilation and the contribution of this document to the selection and refinement of the ADAS&ME Use Case (Chapter 1). An overview of addressed affective states per Use Cases, as they are currently agreed upon, are presented in Chapter 2, along with their general definitions and driver monitoring techniques and systems. Chapter 3 presents an introduction to main research findings per affective state, followed by a literature review overview table, including connection to the project and respective Use Cases. The complete compilation per affective state is included in sub-sections of Annex 1. The table at the end of the Executive Summary includes the number of journals reviewed. In addition, a short literature review was carried out to investigate the automation strategies for stabilisation mechanisms for motorcycles. The work and relevant findings are included in main the document and the rest are also in an Annex. The literatures review is followed by the benchmarking of existing technologies (SoA) for drivers and riders per addressed affective state (Chapter 4). Similarly, an overview of findings is presented in the table at the end of the summary. In addition, studies from other transport modes were reviewed, for their potential of transferring knowledge, methods and techniques. The results are summarised in a Table in Section 4.3, but the whole compilation of the studies was annexed. Chapter 5 presents the alignment of A1.1 results with the efforts of several established organizations, such as ACEM, ERTRAC, NCAP and the international collaborations within CARTRE project. The Deliverable concludes with Chapter 6, where an overview of the results, in terms of numbers and relevance (key categorization) to the project, are presented.

Abstract

The process of developing new automotive systems includes various testing cycles to assure a save operation in traffic. Physical system testing on test tracks is very important for this purpose, but rather expensive and might only become possible in later stages of the development process. Using a virtual simulation environment offers a safe possibility of testing new systems in early stages of development. Aditionally, driver-in-the-loop tests at test track and in a virtual simulator make it possible to evaluate driver reaction and potential acceptance by the future users of those systems. Within PROSPECT the new functions are investigated under various aspects in several simulator studies and test track studies. This deliverable D7.3 gives detailed information of conduction and results of the each study. Three studies focus exclusively on the for Vulnerable Road Users (VRUs) specifically dangerous urban intersection scenarios. The first of those studies examines the driver behaviour in a turning situation when a byciclist might be crossing. The described phenomena are looked-but-failed-to-see and failed-to-look. The second study, which provides an initial step in this line of research, analyzed the acceptance of issued forward collision warning times. The positioning of the potential accident opponent and the subjective feeling towards the criticality of the situation by the driver were key parameters taken into account. Last, but not least the acceptance of an intersection assist autonomous emergency braking systems was tested regarding the acceptance of potential buyers. The study was run for five days in a row for each participant to be able to judge the behaviour in a comuting situation. Two studies focused on longitudinal scenarios. Both studies followed the same design, but one was conducted on a test track and the other one in a simulator. The main objective was to investigate drivers reactions to FCW warnings and Active Steering interventions in critical VRU scenarios in case of a distraction of the driver. Additionally, the test track study was used to validate the results from the simulator study. The results of those studies are the basis for a wide acceptance evaluation of the systems. No system is an asset in increasing road safety if it is not accepted by the user and therefore turned off, if it is not required the system to be default on in consumer tests. Complemented by an additional acceptance study where the participants had to give their opinion of those systems after they watched videos of dangerous situations, the acceptance was analyzed based on questionnaires developed in PROSPECT and reported in Deliverable 7.2. This wholistic approach allows an expert discussion on the potentials of the PROSPECT functions in the future.

Abstract

The aim of Task 5.1 is to identify and select, among the outcomes of SP4, the results that are relevant to infer recommendations for measures improving road safety and sustainability. Due to time constraint, the analyses and the recommendations have been done in less time that it was planned at the beginning of the project. The key outcomes of the SP4 work with particular reference to crash risk, unsafe driving, and eco-driving will be studied and organized in terms of relevance to safety and sustainability policies and potential actions towards road users, vehicle and road. Recommendations have been developed to propose actions to stakeholders that can be implemented in the near future to increase safety and sustainability of road transport. This work integrates several reviews of different measures implemented previously in France, Germany, Netherlands and United Kingdom in terms of road safety measures. Then, the recommendations consider possible updates of existing measures and the development of new measures.

Abstract

Deliverable D2.1 'Accident Analysis, Naturalistic Observations and Project Implications' is issued in the scope of WP2 'Accident analysis and user needs' from the PROSPECT project. The objective of WP2 is to generate the user requirements for next generation proactive safety systems, with a focus on the specific needs of vulnerable road users (VRUs). Part A of deliverable D2.1 (Accident data analyses) provided results from task T2.1 'Characteristics of vehicle to VRU accidents'. Within this task, an in-depth accident analysis involving Vulnerable Road Users was carried out in Europe, focusing mainly on pedestrians and cyclists.The output already obtained from task T2.1 has provided information about the current safety situation and the identification of the most relevant car-to-cyclist and car-to-pedestrian accident scenarios where safety improvements are necessary. This data has been used to define the use cases of PROSPECT, and the system development will focus on the most relevant of these. The overall process of use case definition for PROSPECT and the associated test catalogue derived from the accident analysis data is provided in deliverable D3.1 'The addressed VRU scenarios within PROSPECT and associated test catalogue', available in May 2016.This report corresponds to Part B of deliverable D2.1, which seeks to provide additional knowledge to the project through naturalistic observations within selected European cities in order to establish how vehicles and VRUs interact in real traffic situations. This work has been developed in task T2.2.Naturalistic observations facilitate a better understanding of potentially dangerous traffic situations with VRUs. In particular, it includes the identification of motions, behaviours and interactions that lead to such situations, from both VRU and driver perspective. Additional to the information provided from the accident databases, it is necessary to identify the parameters that signal VRU intent in order to enable earlier and more precise reactions by safety systems. Naturalistic observations are therefore crucial for the development of advanced algorithms integrated in next generation PROSPECT-like systems, and must be also taken into account as relevant factors for the definition of test scenarios.An introduction and specific objectives of the task are presented in this part, as well as the methodology for data acquisition and extraction of conflicts regarding VRUs in real-world traffic from infrastructure-mounted and/or vehicle-based sensors and cameras in Lyon, Budapest and Barcelona. An additional study made on Helmond on cyclist behaviour is also described.The parameters considered for the analyses of conflicts are provided, as well as analysis of the conflicts.Finally, this part of the document offers a general conclusion about the results obtained from the naturalistic observations.

Abstract

International audience; Semi-autonomous and fully-autonomous systems must have knowledge about the objects in their environment to ensure a safe navigation. Modern approaches implement deep learning techniques to train a neural network for object detection. This project will study the effectiveness of using several promising algorithms such as Faster R-CNN, SSD, and different versions of YOLO, to detect, classify, and track objects in near real-time fluvial domain. Since no dataset is available for this purpose in literature, we first started by annotating a dataset of 2488 images with almost 35 400 annotations for training the convolutional neural network architectures. We made this data set openly accessible for the community working on this area. The other contribution of this research is the adaptation and the configuration of deep learning techniques used in other domains such as maritime and road domain to fluvial domain for autonomous vessels in which high accuracy and fast processing are vital. Experiments demonstrated that detecting objects in such environment is plausible in near real time with the selected algorithms.

Abstract

International audience; This paper presents a general framework to assess urban rail logistics suitability via a socio-economic cost benefit analysis and demand catchment assessment. First, we examine the main background issues related to data production in urban logistics and the needs of examining the economic viability of future projects. Second, we present methodology proposed that combines a classical CBA framework with a freight catchment analysis framework. Third, as an example of application, we propose to assess freight tramway deployment scenarios in Paris, France. The results show the potential of including demand catchment analysis in CBA to identify the suitable demand introduction strategy. However, it seems important to contextualize them, such methods being used to support decisions and not to substitute decision maker.

Abstract

International audience; Problems tackled by researchers and data scientists in aviation and air traffic management (ATM) require manipulating large amounts of data representing trajectories, flight parameters and geographical descriptions of the airspace they fly through. The traffic library for the Python programming language defines an interface to usual processing and data analysis methods to be applied on aircraft trajectories and airspaces. This paper presents how traffic accesses different sources of data, leverages processing methods to clean, filter, clip or resample trajectories, and compares trajectory clustering methods on a sample dataset of trajectories above Switzerland.