E11
Autonomous Things 2

Back to overview

10:40
conference time (CEST, Berlin)
Co-simulation Framework for Virtual V&V of GNC Algorithms for Autonomous UAV
27/10/2021 10:40 conference time (CEST, Berlin)
Room: E
V. Dezobry, F. Cappuzzo, E. Carencotte (Siemens Digital Industries Software, FRA); S. Di Gennaro; D. Bianchi (Università degli Studi dell'Aquila, ITA)
V. Dezobry, F. Cappuzzo, E. Carencotte (Siemens Digital Industries Software, FRA); S. Di Gennaro; D. Bianchi (Università degli Studi dell'Aquila, ITA)
Unmanned aerial vehicles (UAVs) are disrupting multiples industries such as logistic, agriculture, construction, etc. This broad spectrum of applications requires drone manufacturers to develop a modular platform adaptable to mission needs. This customization applies on the hardware components as well as on the software components and more specifically on the guidance, navigation and controls (GNC). The software development lifecycle goes through a series of stages, from requirement analysis to maintenance. Among these, the verification and validation (V&V) consists of checking that a software system meets requirements and specifications. Testing navigation algorithms with real UAV platforms is an expensive and time-consuming process. Using a simulation tool to test such algorithms provides an alternative to these problems. This paper presents the technique of using a system simulation tool to test control algorithms of an autonomous octocopter UAV, with four coaxial contra-rotating propellers, used in the frame of the European research project COMP4DRONES. The software Simcenter Amesim, a multi-physics system simulation tool, is used to model the different subsystems of the UAV: batteries, propulsion chain and flight dynamics. Those plant models have been favorably compared against experimental data provided by the manufacturers first. Next the overall performance model has been validated with respect to experimental data coming from flight tests. The overall plant model was then integrated in a co-simulation framework capable of modelling drone’s navigation sensors (camera, LIDAR, etc.), mission environment and GNC algorithms to simulate the drone’s behavior under different scenarios: precision landing maneuvers, obstacles avoidance and cluster flight. This framework enables UAV integrators to conduct exhaustive flight tests, easily change the environment by adding more obstacles, perform extreme tests and assess the impact on the drone stability in case of failure. Running GNC algorithms virtual validation ensure the drone behaves properly in multiple environments and conditions, and consequently improve the product performance.
UAV, GNC, simulation, validation, virtual, autonomous
11:00
conference time (CEST, Berlin)
Improvement of Lidar Detection and Tracking Algorithm Using and Development of a Multi-sensor Fusion Module
27/10/2021 11:00 conference time (CEST, Berlin)
Room: E
M. Hadj-Bachir, D. Gruyer (Université Gustave Eiffel, FRA); K. Fcheris Kevin, P. de Souza (ESI Group, FRA)
M. Hadj-Bachir, D. Gruyer (Université Gustave Eiffel, FRA); K. Fcheris Kevin, P. de Souza (ESI Group, FRA)
Ensuring the reliability of the perception of an intelligent vehicle is recognized as one of the challenges for the transition to the higher levels of autonomous driving. From a technological perspective, safety is enforced using combinations of sensors systems exploiting different physics principles used by optical sensor such as camera [1], by near infrared sensors like laser scanner [2], electromagnetic sensors like RADAR [3] and radiofrequency and ultrasonic sensors. Each sensing system is defined by its own hardware components, electronics and detection function with embedded processing which deliver a variable performance according the driving environment characteristics. Validating of perception functions for autonomous driving– and of the sub-functions involving the detection, tracking, object recognition– is extremely complex by the extent and the variety of the conditions to be tested (sensor, infrastructure, object, weather, … features). Gradually, the industry has embraced the idea of introducing simulation methods to support such validation [4, 5]. This study aims to improve a targets detection and tracking algorithm by laser scanner. These improvements relate to the targets detection and tracking modules in which new functionalities have been implemented to improve algorithm results. A code optimization has been implemented to improve the execution time of the algorithm and thereby increase the performance of data processing. The algorithm was tested with real data of SICK LIDAR, and synthetic LIDAR data with 4 and 16 layers simulated by PROSIVIC platform. The use of simulation helps us to study the influence of several parameters on the detection like effects of impacts points number, weather condition and to test the algorithm robustness. In addition, a study of the libraries was carried out in order to know the dependencies of the algorithm for the power to integrate at best in the PROSIVIC simulation platform. This report thus made it possible to show how these improvements were implemented in order to achieve the objectives as well as the functioning of these improvements
ADAS, Autonomous vehicles, sensors, LIDAR, RADAR, camera, Ultrasonic, GPS, NCAP, NHTSA, post-processing, tracking, data fusion, interoperability
11:20
conference time (CEST, Berlin)
Simulation of ADAS Functions in Cutoff Situations
27/10/2021 11:20 conference time (CEST, Berlin)
Room: E
M. Kereszter, K. Pintér (Bay Zoltán Nonprofit Ltd. for Applied Research, HUN)
M. Kereszter, K. Pintér (Bay Zoltán Nonprofit Ltd. for Applied Research, HUN)
The laboratory of the Bay Zoltán Nonprofit Ltd. located at the ZalaZONE proving gound has the goal to test and develop Automated Driving and Advanced Driver Assistance Systems applications applying a Driver-in-the-Loop Simulator. In both topics we perform simulations and support the preparations of physical tests, further on we have our research and development topics. In our recent study we focus on to identify bottlenecks of the trending collision avoidance technology, which is one of the fastest growing ADAS feature field. Nowadays we have many different ADAS solutions which are working well on their own. Our main task now is to develop an intelligent autonomous driving algorythm for vehicles which is able to choose the best option in a critical situation to avoid a possible accident or at least mitigate damages. For this it is really important to have the right informations of the vehicle just like vehicle acceleration or wheel speed. We have to define the last segment of the trajectory where the vehicle still has a chance to make its decision about the best maneuver. For the investigations we choose two ADAS solutions, an Automatic Emergency Braking (AEB) system as well as an Emergency Steering Assist (ESA) system and our intention is to combine their operation. In a critical situation the boundary conditions are changing constantly in every miliseconds. The investigated situation is a standard motorway case, where the Vehicle-under-Test (VUT) is following the Global Vehicle Target (GVT), with the minimal sufficient tracking distance. We investigate this scenario where the GVT starts an emergency braking and this triggers the VUT to do the same. This maneuver generates a critical situation where we investigate the VUT’s opportunities beneath different environmental factors. First we will decrease the coefficient of friction of the road surface what will result the increase of the braking distance. This way applying only emergency braking will become inadequate the avoid the collision and the VUT has to be moved in lateral direction, too. For these scenarios we are using automotive simulation software where we can implement and simulate different sensors on the vehicle and we can test them effectively applying different parameter sets.
Advanced Driver Assistance Systems, simulation, automotive, autonomous driving
11:40
conference time (CEST, Berlin)
Surrogate Model Based Safety Performance Assessment of Integrated Vehicle Safety Systems
27/10/2021 11:40 conference time (CEST, Berlin)
Room: E
P. Wimmer, S. Kirschbichler, O. Zehbe (Virtual Vehicle Research GmbH, AUT); J. Hay, L. Schories (ZF Friedrichshafen Corporate Research and Development, DEU); J. Fehr (University of Stuttgart, DEU); E. Bayerschen, (ZF Friedrichshafen Passive Safety System
P. Wimmer, S. Kirschbichler, O. Zehbe (Virtual Vehicle Research GmbH, AUT); J. Hay, L. Schories (ZF Friedrichshafen Corporate Research and Development, DEU); J. Fehr (University of Stuttgart, DEU); E. Bayerschen, (ZF Friedrichshafen Passive Safety System
The future of vehicle safety will bring a fusion of active and passive safety into integrated safety to further reduce the number of injuries in road traffic. This development requires a closer interaction of different simulation domains that are nowadays separated. Therefore, new methods and processes must be established so that integrated safety systems can be developed and assessed accordingly. These methods and processes must cover relevant aspects from normal driving to crash, e.g. driving dynamics, sensors, active-safety algorithms, vehicle structure deformations, restraint systems, and occupant behaviour. Apart from the need to combine all these different simulation domains, this development poses an additional challenge to passive safety system evaluation: A scenario-based assessment consisting of a large number of simulation runs instead of evaluating a limited number of test cases. Classical finite element crash simulations require considerable simulation effort to deliver precise results and are therefore not suited for such types of large-scale analysis. Consequently, new, time-efficient methods need to be developed , which significantly reduce calculation time while still providing acceptable result prediction quality. As a possible solution addressing this challenge, mathematical and physical surrogate models for all time-consuming simulation steps are presented in this paper, enabling a time-efficient safety performance assessment of combined active and passive safety systems. As a proof-of-concept of the method, an assessment of two integrated safety system variants consisting of an autonomous braking (AEB) system and a standard restraint system is presented. The potential of the integrated safety system to reduce occupant injury risk is shown using 285 virtually generated accidents, resulting in a total number of 285*3=855 simulation runs (285 runs for the baseline plus 285 runs for each integrated safety system variant). As an additional benefit, the developed fast-calculating surrogate models’ usage also allows for a case-specific optimisation of the restraint system. By that, the potential for reducing the occupant injury risk can be increased even more.
Safety performance assessment, integrated vehicle safety, continuous framework, surrogate models
×

[TITLE]

[LISTING

[ABSTRACT]

[DATE]

[ROOM]

[KEYWORDS]