Skip to main content
Erschienen in:

Open Access 2022 | OriginalPaper | Buchkapitel

7. Summary and Discussion

Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.

search-config
download
DOWNLOAD
print
DRUCKEN
insite
SUCHEN
loading …

Abstract

The automotive industry is in the progress of a fundamental change, as they no longer meet mobility requirements, especially in urban areas. As a result, many predict a disruptive change. Responses to this are new innovative developments. One answer to this are automated driving systems that offer great potential for increasing safety, comfort, environmental pollution and efficiency in road traffic.

7.1 Current agile management changes

From the author's observation, increasing economic pressure is requiring leadership and employee teams to become more innovative, faster, efficient, highly profitable and more prepared to take risks. It is no longer sufficient to work effectively or goal-oriented.
This will give rise to concerns that new innovative developments using Artificial Intelligence could lead to external control by Artificial Intelligence. This focuses on ensuring controllable collaboration between humans and increasingly intelligent machines. Approaches to the challenges of new developments and development teams in the dilemma between Artificial Intelligence, ethics and the legal risks are considered in more detail here using the example of Automated Driving.
The automotive industry, for example, is in the progress of a fundamental change, as they no longer meet mobility requirements, especially in urban areas. As a result, many predict a disruptive change. Responses to this are new innovative developments (Dudenhöffer, 2016). One answer to this are automated driving systems that offer great potential for increasing safety, comfort, environmental pollution and efficiency in road traffic. In the long term, fully automated or autonomous vehicles offer many useful advantages: While driving, non-driving activities can be done and so this time is used efficiently. Older or physically handicapped people can also become mobile again. It also supports new business areas, especially in the area of car sharing.
The business models differ radically. The approaches of Google, Apple, Facebook or Tesla do not aim for profit margins from the sale of automobiles, but for security and expansion of data competence, a new level of networking. The customer of the future is increasingly looking for new mobility services to get from A to B.
Until now, the German road traffic regulations (Straßenverkehrsordnung StVO) had firmly established the permanent controllability of a vehicle. According to § 3 Section 1 StVO, for example, the driver may only drive so fast that he is able to control the vehicle at all times. Initiatives at UN level are continuing to drive forward the “ALKS - Automated Lane Keeping System” with future speed extensions. Similarly, the German Bundesrat has passed a law on autonomous driving in 2021. As described in section 4.​7.​1, some cases are already known and published, where unexpected or missing reactions of automated systems occurred. However, after fatal traffic accidents (for example Tesla „Autopilot“ 2016/2018, Uber self-driving vehicle, 2018) automated vehicles must face the discussion of the dilemma between innovation and consumer protection, which leads to a deeper need for research according to the Requirements to Develop Safe Automated Vehicles.
Previous safety methods will no longer be sufficient for the verification of complex automated driving functions. Therefore, the requirements to develop safe automated vehicles between the dilemma of innovation and consumer protection were examined in more detail in this study.
The following topics have been processed:
  • Existing development specifications for use in the development of partially, highly and fully automated vehicles (see chapters 4, 4.​6)
  • Instruments to ensure the required quality of the safety process of automated vehicles (see chapters 2, 3, 5, 6)
  • Expectations of potential users and developers for the product safety of automated vehicles (see chapters 2, 4, 4.​5)
  • Increasing the product safety of automated vehicles by taking expert experience into account (see chapters 4.​7, 5)

7.2 Findings

This book demonstrates that both professional knowledge with sustainable leadership decisions and concepts of psychological development with mindfulness are growing in importance. It further confirms that development work using the example of automated driving functions with Artificial Intelligence will be successful through close structured support and a sustainable Leadership between teams of experts. Such close teamwork including the knowledge from area-wide accident data in addition to other field studies (Driving Simulator, Natural Driving Studies, Field Operational Studies), legal framework conditions with liability cases and validation methods will support the development of safe automated vehicles. In the future, the main focus will be on developing the level of safety that automobile manufacturers have to ensure. Finally, court decisions will decide on the permitted risk in concrete cases. A definition of a permitted risk would be suitable to structure and limit the criminal liability of manufacturers of automated systems appropriately in the future.
The usage of the final consulting concept (including feedback from the development departments and the checklist in Annex B) developed in this book is a way to reduce the risk of criminal consequences for the company plus the threat of prison punishments for individual employees. The concept supports the development of an automated vehicle (in the context of what is technically practicable) as safely as possible.
With the support of this checklist concept, the developers have resources and common understanding to reduce criminal consequences to an absolute minimum. It demonstrates that the most appropriate procedures have been applied in development, including risk identification, risk assessment, and assessment methodology.
Initially, the exemplary findings of traffic accident research in chapter 2 indicate that human error – with mainly information reception limits of almost 60 percent – seems to be the main cause of road accidents. In the first instance, this raises great expectations for the benefit of automated vehicles. However, estimating the actual safety potential of highly and fully automated vehicles from accident data, therefore, requires a differentiated comparison of the overall performance of man and machine. Subsequently, this calls for detailed information about functional characteristics and technical limits, planned for mass production.
Before series development is considered, driverless vehicles, supported by automated systems, must at least correspond to the driving ability of an attentive human driver to further reduce the number of road accidents.
For example, development engineers are particularly faced with technical challenges regarding complex traffic situations. This applies, for instance, to technical limits and time-critical situations, such as a child running suddenly in front of a vehicle or difficult weather conditions. Only when these technical challenges have been overcome is a large-scale rollout of marketable, fully automatic vehicles likely to be realized.
The potential of information from traffic accident data is not yet completely used. Previous accident analyses are usually not nationwide and limited by criteria. Predefined analysis criteria of accident research teams are usually limited to certain locations, times, special collision conditions (such as airbag deployment, involvement of injured persons, special pedestrian accidents, vehicle types or other general conditions) and must therefore first be weighted for statistical relevance. For example, area-wide minor accidents with minimal contact and minimal damage to property (see Sect. 3.​3.​2.​5 Examples for minor and no damage to property), or traffic violations that come very close to “near misses” are not investigated in depth.
To receive real-world test scenarios for the first time 1,286,109 state-wide police-recorded accidents were analyzed concerning challenging information reception limits in chapter 3. The results indicate 374 scenarios with bad weather traffic conditions (fog, glare/blinding sun, rain, black ice (snow/ice), snowfall, blinding oncoming traffic, visual obstruction) that are also relevant for testing automotive sensor systems.
In particular, a fatal pedestrian accident scene was examined at the accident site under similar conditions concerning the perception capabilities in comparison of human and machine. The situation shows that such indicated scenarios have to be considered for sign-off testing after the careful selection of sensor concepts and the development of algorithms.
Consumers require the highest, state of the art levels of safety for new technologies but those demands for technical perfection are unrealistic and 100% fault-free operation is not possible. A market introduction of automated vehicles accompanies the risk that court decisions will be passed more frequently to design faults since a certain risk of accidents can never be completely excluded. However, the liability of the manufacturer is excluded if the defect could not be detected according to the state of science and technology at the time when the manufacturer placed the product on the market. The manufacturers are obliged to observe their products. This can be supported by the analysis of accident events as described in chapters 2 and 3.
Thus, the results of chapter 4 show, that interdisciplinary coordinated development, and sign-off processes are necessary. A reliable evaluation for production-ready solutions requires comparable risk assessments and safety proofs, e.g. by simulating relevant scenarios including the planning of field tests from globally available and combined accident, traffic flow, weather and vehicle operating data.
This also covers compliance with legal and licensing regulations, the identification of new ways of risk distribution and the creation of new compensation systems, because, with the increasing use of automated vehicles, the manufacturer’s liability may also increase. Today the standards do not cover functional disabilities for instance misinterpretation of objects, traffic situations and resulting false positive system interventions.
Qualitative interviews in development departments of German automobile manufacturers show that structural, legal and regulatory support by independent experts in conjunction with sustainable leadership as well as a guideline-based structure can make a significant contribution to the safe development of new innovative systems using Artificial Intelligence. The results of this survey in chapter 5 show that the main challenge for the employees of the development departments is to develop these new systems in a customer-oriented, safe and controllable manner for the vehicle users:
It turns out that engineers are looking for meaningful creativity although they work under tremendous time pressure when developing a new system. In contrast to that, executives are primarily focused on the responsibility for liability and a timely sign-off. They expect the fulfillment of the safety requirements and a completely documented development process. This is presumed because they are afraid to be sued for dangerous situations and accidents due to technical system limits or operating errors at the end-user, which can also lead to a painful loss of image for the manufacturer.
In particular, the survey showed that a structured guideline with supporting advice forces the parties to come together on an interdisciplinary basis, to clearly present and discuss their diverging aspects and to decide according to the duty of care.
One effect of the survey was that it sensitized the interviewed development departments to the advantages of a guideline-based development process. The interviews also show that usually, the developers themselves with their technical expertise develop safe automated vehicle systems when they are motivated to engage in interdisciplinary exchange with other experts from neighboring disciplines. Design engineers know the weaknesses of their new technical system best and can initiate innovations “bottom-up” in companies.
Additionally, the interviews confirm that a guideline-based approach enables the affected developers to clearly and neutrally point out risks with corresponding proposals for measures because they know the limitations of their new technical system best.
A selection of 303 key questions (Annex B) from technical requirements, sustainability, Corporate Social Responsibility, up to criteria for communication, team work, leadership qualifications as well as Inner Development Goals supplemented by a consulting concept in chapter 6 supports the establishment of standardized processes and consulting-services.

7.3 Integration of findings

A mindful corporate team management, within new working conditions (New Work, Hybrid Work, Flexible Work) including Artificial Intelligence, must show a sustainable orientation for new developments. A sustainable orientation includes knowledge of the existing objective facts, the possibilities and risks of Artificial Intelligence, as well as the ethical and legal implications. According to this, political, economic and social decisions must be aligned in an equally sustainable manner. Artificial Intelligence can support managers in linking knowledge and designing complex functions, such as automated driving. In this context, management must ensure that new innovative product ideas – with the help of comprehensive information on secured development methods, production and release processes, through to marketing and product monitoring – are carried out with the intention of humane use.
This book also demonstrates using the example of automated driving that successfully interacting teams can reduce the criminal consequences for the company and the individual developer to a minimum if guideline-based checklists with relevant standards and methods are applied. By integrating the findings, it will be supported in dealing adequately with the new challenges facing automobile manufacturers and their developers in the field of functional safety of complex electrical/electronic systems and software topics to prevent from the criminal law punishments of a “defective product”:
1. Successfully self-reflected teams benefit from a guideline-based work process including requirements and consumer expectations:
The basic prerequisite for successful teams are self-reflective team members who work together successfully (see checklist in Annex B from question 102 with inner development goals). Only in a successfully interacting team of experts Original Equipment Manufacturers can cope with new technical developments, legal developments on product liability, international conflicts, ethical and social requirements in connection with economic risks. This means that very high demands for quality and safety are placed on the development from product idea to marketing, whereby customer expectations on the functionality and safety of use with a correspondingly strong influence on traffic safety are of primary importance. Events in recent years have publicly shown that failure to comply with specifications can result in legal responsibility for developers and executives.
Predictions according to the ADAS Code of Practice and current questionnaires of more than 3000 people in Germany, the USA and China confirm that the expectations for functional safety are rising with an increasing level of automation (see Annex Fig. A.2 and A.4 to A.6). Therefore, an extension of the established test procedures is necessary to enable automated driving levels and at the same time to consider the entire range of possible traffic situations as comprehensively as possible in the safety tests.
For the development process from the first idea to the development, this elaboration recommends interdisciplinary, harmonized safety and test procedures. In this context, the further development of current internationally agreed standards including tools, methodological descriptions, simulations and guidelines with checklists is recommended. These will represent and document the practiced state of science and technology, which must be implemented in a technically suited and economically reasonable way.
2. Successfully development teams benefit from the implementation of comprehensive measures for product monitoring:
Opportunities for product monitoring must be used. This includes, for example, the monitoring of operational data, road accident events, and internet forums. A judgment of the Federal Court of Justice (BGH) as early as 1987 stated that in future companies must not only monitor the reliability of their products in practice, but above all draw their customers’ attention to risks in daily operation – including those arising from the use or installation of accessories from other manufacturers.
The potential of information from nationwide databases and traffic accident data is far from fully explored. Previous accident analyses are mostly limited by criteria. Certainly, traffic accidents only represent a part of the traffic situation, but they play an important role in terms of consumer protection with civil and criminal law implications. Furthermore, small accidents with minor contact come very close to “near-accidents”. An analysis of traffic violations, which has not been discussed here, could also provide valuable information.
For the development and validation of safe automated vehicles with reasonable effort, the author recommends test methods that consider a combination of worldwide traffic accidents, weather-, vehicle operating data and traffic simulations. This enables a realistic evaluation of internationally prospective traffic scenarios with statistically relevant real traffic scenarios as well as fault processes and stochastic models for controlling critical driving situations. These must be combined with virtual laboratory or driving simulator tests.
A representative driving situation catalogue including challenging and bad weather situations is recommended, which is simulated for all manufacturers according to the same specifications and the results are made available to the official institutions. This procedure ensures transparency of the overall effect of new automated driving functions in real traffic.
When designing driving strategies for behavioral decisions, the focus should be on completely avoiding dilemma situations, for example by designing vehicles for a correspondingly low-risk driving strategy.
3. Successful teams are advised to engage an independent consultant beyond the respective development area:
The interview partners would like to have a neutral face-to-face contact person outside of the development department who is always available to provide competent technical or legal advice in the event of questions and arising problems.
Competent support by an independent consultant from outside the respective area is recommended by all developers being interviewed. This adviser should support decisions and the accompanying documentation during the entire development process in conformity with the duty of care regarding to the central question:
“Is the developed system safe enough for market introduction?”
4. Successful interacting development teams benefit from constant monitoring of social, ethical and legal issues
The author's consulting work within corporate groups shows: Boards of directors and managing executives have recognized that Corporate Social Responsibility (CSR) and sustainable management have a positive impact on corporate success worldwide. The Sustainable Development Goals (SDGs) of the United Nations, including the Inner Development Goals (IDGs), are a valuable contribution to this.
For example, changes in responsibilities regarding road traffic are indicated by the analysis of German court decisions on pedestrian accidents since 2004 (See section 4.​6.​4 and Annex A: Change in jurisdiction on the responsibility for pedestrian accidents). The liability for damages in pedestrian accidents increasingly lies with the owner and therefore in the case of fully automatic functions in the future probably with the manufacturer. As a result, our current risk awareness in road traffic with regard to risk acceptance in automated driving levels must be called into question. An example for this is the child running between parked vehicles. In this context, it must be questioned whether speeds of 50 km/h or more are appropriate in traffic areas with visual obstructions, such as parking vehicles.
Conventional dynamically adapted interactions of today's mobility can also be questioned in terms of whether fully automated vehicles must always behave in accordance with traffic regulations. Today's mobility is based on the fact that in some traffic situations, human pragmatism makes decisions that are weighed up against traffic rules in order to maintain the flow of traffic. An example of this is the continuous road lane marking line that needs to be crossed to overtake a bike or a broken vehicle.
Traffic would probably come to a complete standstill in some places if rules were not broken. Therefore, the challenge is to program the vehicle software in such a way that it considers the illegal behavior of other road users and possibly breaks its own rules to reactivate the traffic flow. This leads to the recommendation that in the future the developers make their ethical decisions regarding the programming of the software within society more transparent, because this is where the opinion is formed which system reactions with corresponding risks are accepted. As long as not all rules for behavioral decisions have been made concerning how automated vehicles should behave in specific situations (when, how, why (or not) warn, steer, brake), the intensive dialogue between developers and system providers with society is recommended. This applies in particular to the performance of Artificial Intelligence self-learning systems. Deeper Neural Networks (DNN) with a depth of more than 150 layers are increasingly easier to optimize today and can improve their precision due to a significantly increased depth with errors of less than 4% in the classification task. As a result, the object recognition data set improves significantly (He K et. al., 2015, 2016).
So far, not all general requirements have been defined as to how a vehicle should behave in specific situations. The discussion about the safe state raises new questions too. Furthermore, it should be mentioned that automation, combined with connected networking, Artificial Intelligence and Deeper Neural Networks, offers new opportunities for cybercrime, another topic that is not discussed in detail here.
The concluding outlook on the current state of science again points to the limits of testability. While trivial systems can be tested, the challenge increases for complex systems. The Department of Motor Vehicles (DMV) in the USA, which is comparable to German road traffic authorities, publishes annual “Disengagement Reports”. This includes, among other things, how often humans had to take corrective action during testing of fully automated vehicles or when the system returned control to the safety driver.
These results indicate on one hand the successful commitment of the Google subcompany Waymo and on the other hand the need to optimize the robustness of fully automated vehicles. While Apple's test drivers had to intervene a total of 871.65 times per 1000 miles traveled (one intervention per 1.1 miles), Waymo’s test drivers only intervened 0.09 interventions per 1000 miles (one intervention every 11,154 miles), (see Annex Fig. A.7).
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://​creativecommons.​org/​licenses/​by/​4.​0/​), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
download
DOWNLOAD
print
DRUCKEN
Metadaten
Titel
Summary and Discussion
verfasst von
Thomas Winkle
Copyright-Jahr
2022
DOI
https://doi.org/10.1007/978-3-658-34293-7_7

    Premium Partner