Guldkorn från svensk forskning

Dessa guldkorn är bidrag från våra läsare – stort tack för det, och för all fantastisk forskning och utveckling som ni gör. Keep up the good work!

iQ-Pilot & iQ-Mobility. These are two recently finished projects co-funded by the Strategic vehicle research and innovation programme (FFI). The focus of the projects was development of new technology to realize flexible, energy-efficient transport solutions in cities. Several proof-of-concept prototypes have been developed and demonstrated, including autonomous buses and a smart coordination system for bus fleets. The research results were presented in a webinar earlier this week. These results are the joint efforts of Scania, Ericsson, INIT, Veoneer, Royal Institute of Technology (KTH) and Örebro University. 

Human interaction with autonomous minibuses. Tom Ziemke’s research group at Linköping University, in collaboration with researchers at VTI, will during the autumn start a new research project on people’s interaction with autonomous minibuses on campus. The research will focus on method development and empirical studies of how pedestrians, bicyclists and car drivers interact with the buses. A two-year postdoc position is available via this link (application deadline: August 5). For more information contact Tom Ziemke (tom.ziemke@liu.se).

GLAD – Goods delivery under the Last mile with Autonomous Driving vehicles. Small autonomous electric delivery vehicles (ADV) are expected to transform transportation of goods under the first and last mile. The advantages are increased transportation and energy effectiveness, but it is also important that these vehicles are safe and accepted in society. The aim of the GLAD project is to develop an initial knowledge base on efficiency, safety and human experience of ADVs for the first and last mile delivery of goods in Sweden, and on how to create a balance between these three aspects from a socio-technical perspective. To achieve this, the project will utilize Zbee vehicles that will be adapted in terms of vehicle design and autonomous vehicle behaviour, human-machine interface, teleoperation and vehicle management. The overall goal is to develop knowledge that accelerate introduction of new efficient goods delivery in our society and contributes to meeting the goals of Agenda 2030. This will be assured also by connecting a licentiate candidate to the project. The project is co-funded by Trafikverket and involves RISE, Halmstad University, Aptiv, Combitech and Clean Motion. It started in June 2020 and will run for ca 2 years. For more information contact azra.habibovic@ri.se.

Tactical Decision-Making in Autonomous Driving by Reinforcement Learning with Uncertainty Estimation. Reinforcement learning (RL) can be used to create a tactical decision-making agent for autonomous driving. However, previous approaches only output decisions and do not provide information about the agent’s confidence in the recommended actions. This paper investigates how a Bayesian RL technique, based on an ensemble of neural networks with additional randomized prior functions (RPF), can be used to estimate the uncertainty of decisions in autonomous driving. A method for classifying whether or not an action should be considered safe is also introduced. The performance of the ensemble RPF method is evaluated by training an agent on a highway driving scenario. It is shown that the trained agent can estimate the uncertainty of its decisions and indicate an unacceptable level when the agent faces a situation that is far from the training distribution. Furthermore, within the training distribution, the ensemble RPF agent outperforms a standard Deep Q-Network agent. In this study, the estimated uncertainty is used to choose safe actions in unknown situations. However, the uncertainty information could also be used to identify situations that should be added to the training process. The paper will be presented at the Intelligent Vehicles Symposium (IV) in October 2020, and a preprint is available on arXiv. The code that was used is also available on GitHub For more information, contact Carl-Johan Hoel (carl-johan.hoel@volvo.com) at Volvo Autonomous solutions. This work was partially supported by the Wallenberg Artificial Intelligence, Autonomous Systems and Software Program (WASP), funded by Knut and Alice Wallenberg Foundation, and partially by Vinnova FFI.

Autonomous Mapping of Unknown Environments Using a UAV. As part of the research conducted within the project LASH-Fire (Eu-Horizon 2020, No.814975), RISE supervised the work of Chalmers students developing an automatic object search for indoor environments using a flying drone. At the core of this system a reinforcement learning (RL) algorithm was implemented for the drone to navigate, detect obstacles, recognize objects and explore the environment. This machine learning (ML) project marks a starting point for further development towards an autonomous identification and surveillance solution in a wide range of study cases where cargo ships, like the ones studied in LASH-Fire, are an ideal target application. A modularized approach was used targeting research areas such as obstacle avoidance, object detection & recognition, simultaneous localization and mapping, etc. The exploration module was specially challenging and will require further work but the project in general was successful in providing a methodology and tools when using flying drones for indoor environments. The Master’s thesis was conducted by Erik Persson and Filip Heikkilä, and is available via this link. For more information contact boris.duran@ri.se

Projektet ESPLANADE, som började 2017 och avslutades sista mars 2020, handlade om hur man visar att ett automatiserat fordon är säkert. Det finns flera problem som måste hanteras för att man ska kunna göra en komplett säkerhetsargumentation. Projektets resultat inkluderar därför nya metoder för säkerhetsargumentation för en ADS, några av dessa är: 

  • En process för säkerhetsanalys samt designprinciper för interaktionen när en människa överlämnar kontrollen över ett fordon till en ADS eller tvärtom. Processen innehåller existerande metoder som sekvensdiagram, orsak-konsekvensanalys och felträd, men applicerade på människa-maskininteraktion istället för enbart tekniska system
  •  Hur man definierar den operativa designdomänen (ODD) för en ADS utgående från önskade användningsfall, vilket innebär en definition av parametrar inom vilka en ADS-funktion är avsedd att fungera, samt strategier för att säkerställa att fordonet håller sig inom sin ODD.
  • En metod (kallad QRN) för riskanalys och framtagande av säkerhetsmål. Till skillnad från vanliga riskanalysmetoder bygger den inte på analys av specifika situationer utan på definition av acceptabel frekvens av incidenter med olika allvarlig konsekvens, och en mappning av incidenter till olika klasser av konsekvenser. Säkerhetsmålen uttrycks så att man säkert hamnar inom acceptabla frekvenser.
  • Ett ramverk för formell och systematisk hantering av säkerhetskrav med en kombination av åtgärder under utveckling och under drift, bland annat baserat på modeller av osäkerhet.
  • Användning av metoden funktionsanalys för att distribuera beslutsfattande på en ADS-arkitektur samt framtagande av säkerhetskrav.
  • Säkerhetskontrakt och komponentbaserad design för att underlätta kompletthetsbevisning i kravnedbrytning, möjliggöra kontinuerlig produktuppdatering, samt kunna uttrycka säkerhetskrav för sensorsystem som inkluderar kamera, radar mm.

En publik rapport och länkar till de flesta av projektets publikationer finns på projekthemsidan.

Prepare Ships Project. Running for 26 months, the H2020 project “Prepare Ships”, funded by the European Global Navigation Satellite System Agency (GSA), was successfully started in December 2019. The 5 consortium partners, coming from 3 European countries have developed a machine learning based future position prediction for ships in order to avoid ship collisions and close quarter situations as well as reducing environmental impact by more advanced decision making. In a RTK (Real Time Kinematic) software solution, it will both exploiting the distinguished features of Galileo signals as well as combining it with other positioning and sensor technologies. It will use the next generation maritime communication techniques VDES and the new suit of IALA Standards (S100) on sea charts. The innovation developed during the project can make more autonomy of navigation feasible by exchanging future positions and allow eased decision making on ships, suitable to become an international game changer for the future of autonomous shipping. The demonstration and testing will be done onboard three different vessels in the Gothenburg archipelago. The project is coordinated by RISE with partners from across Europe, including SAAB, Lantmäteriet, Telko and Anavs. For more information check out our homepage, join our linkedin group or contact Johannes Hüffmeier at RISE (johannes.huffmeier@ri.se).  

How do you ensure safety of autonomous shipping? Today’s risk assessment methods, application of methods and models used in shipping are usually based on humans being directly in charge of ships, VTS, port controls, etc. and may not be sufficient to reflect and evaluate the complexities and inherent risks of introducing further automation and digitalization in the shipping domain. The introduction of smart ships will create traffic situations between manned and unmanned ships where on one hand decisions and actions are based on algorithms and on the other hand by a human operator where a large part of the decision making. Increasing the level of automation implies that the goal-based standards for shipping need to be based on a risk assessment that reflects the expected roadmaps towards more smart ships and so far, research on autonomous transportation has focused on other parts than the effect of introducing and mixing different levels of automation and only very basic standards have been proposed by classification societies, where DNVs standards [DNV, 2018] have two pages in the appendix on basic set-ups for testing and validation. The main objective of the RFAF project financed by Trafikverket is to analyse how autonomous navigation can be proven to be safe. The aim of the project is to perform a simulator-based risk identification for autonomous shipping traffic. Increasing the level of automation implies that the goal-based standards for shipping need to be based on a risk assessment that reflects the expected roadmaps towards autonomy. Based on two use cases, the routes Fredrikshamn-Göteborg and crossing of the Ljusterö fairway, relevant risks are identified based on ship simulations performed by mariners describing especially nautical challenges for more autonomous shipping resulting in a common risk model. The project lasts from January 2020-December 2022. There are 3 project partners with RISE as coordinator. For more information visit the project website or contact Johannes Hüffmeier (johannes.huffmeier@ri.se).

The SWEA-financed (Energimyndigheten) Data-driven Optimised Energy Efficiency of Ships is a national project involving 7 ship owners, 3 companies from the supply chain and RISE, lasting for 16 months. The data analysis of energy consumption is often complex and there are different driving forces for decisions. However, increased data collection can be unprofitable if you do not have methods to analyze the complex systems. Developments within machine learning provides new opportunities to develop both technically and economically powerful tools energy efficiency. Even today, to some extent, economic driving is applied, for example. eco-driving, however, the effect is in many cases limited as decision-making is more complex than the operator / navigator can see. Also, not always available incentives and motivation of individuals to reduce energy use. However, data collection is increasing both quality review and analysis are not performed to the same extent. Using the results of the project’s data collection and analysis, recommendations can be given about which tools which can be developed in a next step, such as: a) nudging, decision support system or autopilot for ECO driving, b) route optimization based on the ship’s accelerations and motions, and c) decision support based on statistics or real-time analysis of data to identify optimal operation (parameters such as sea state, current, speed, load condition, etc.). The objectives of the project are to: a) Achieve reduced energy use on the project’s vessels by 10–35% both at quay and in sea operations, b) Demonstrate potential with machine learning of operational data, and c) Demonstrate the possibility that better operational data may form the basis for the development of generic energy efficiency tools for smaller vessels in commercial traffic. For any details on the project, reach out to Johannes Hüffmeier (johannes.huffmeier@ri.se).

Photonics Private Public Partnership Roadmaps for EU’s next Framework Program Horizon EuropéThe area of photonics for automotive applications is a significant area which includes not only photonics sensors for the EU defined topic Mobility and Safety for automated Road Transport. Photonics also plays a role in the path towards the targets of Zero Emission Road Transport, Clean Energy Transition, and the Industrial Battery Value Chain. The work of defining the Strategic Research Agenda (SRA) in the specific area of Photonics with EU industrial partners, universities and research centers is performed through the EU technology platform ”Photonics21”, which is funded by the EU commission. The current roadmap for Photonics was published in the document: “Europe’s age of light! How photonics will power growth and innovation, Strategic Roadmap 2021–2027” The section on Automotive and Transport can be found in section 3.9. The coordinator of the whole Photonics 21 is done by VDI Technologiezentrum GmbH in Düsseldorf, Link. We believe this is important as there are a lot of EU research money at stake. The current recommendation by the European Parliament for the whole Horizon Europe budget 2021 -2017 is €120 Billion. The research funding will be divided among many topics where Climate, Energy, and Mobility is one of the clusters. There is a large Swedish interest in the cluster and cooperation with industry is one important factor in the program. Most, if not all, of the European automotive industry are usually involved in at least selected programs.

Now, based on feedback from the new European Commission, the board of Photonic21 have decided to reshape the roadmap and as a consequence automotive & transport will henceforth be combined with the topics of climate and energy. Besides merging the different topics in one document, this gives us an opportunity to revise the previous document into something that we believe should support our industry even better, considering that the current document was prepared in 2018 and the present situation the industry is facing. We want to ensure that the guiding document capture the specific needs of the automotive industry. The aim of the work is to define the research topics of the Strategic Research Agenda (SRA) which will define the upcoming calls in the Horizon Europe program. 

We now invite comments on the current chapter and roadmap (provided in the link above). Determined by EU commission schedules this work has to be completed on 4 September, why we need your input no later than 24 August 2020. We ask for specific text suggestions and specific roadmap suggestions (compare with p. 140 in the above mentioned Strategic Roadmap). Please forward your suggestions to Jan-Erik Källhammer at jan-erik.kallhammer@veoneer.com. He acted as chair of the group Automotive and Transport in the current roadmap and now act as co-chair of the new group Climate, Energy, and Mobility together with Dr. Heinz Seyringer of V-Research GmbH in Austria.