安全駕駛倡議：創(chuàng)建安全自主車輛政策

1、Cover: Getty Images/MetamorworksContentsPreface1 Executive summary62 Key terminology62.1 Defining verification methods2.2 Other reference terms3 Motivation73.1 State of the industry73.2 Current state of autonomous vehicle policy initiatives9Technical standards9International validation initiatives104

2、 AV policy case studies104.1 USA self-certificationCase insights4.2 United Kingdom code of practiceCase insights4.3 Singapore operational safety assessmentCase insights4.4 Discussion5 Synthesis5.1 Opportunities5.2 Addressing the safety gap in AV policy5.3 Objective of SafeDI frameworkContributorsEnd

3、notes 2020 World Economic Forum. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, including photocopying and recording, or by any information storage and retrieval system.PrefaceMichelle Avary,Head of Automotiveand Autonomous Mobility, Wo

4、rld Economic ForumTim Dawkins,Project Lead, Automotive and Autonomous Mobility, World Economic ForumA cohesive approach to autonomous vehicles is possible only when accompanied bya clear top-level vision for the future mobility ecosystem.Autonomous vehicles (AVs) promise to revolutionize the future

5、of mobility, from how we travel and commute to automating the operations of logistics and supply chains. While this promises a future of efficient transport with fewer road fatalities and injuries, the AV technologies remain under development, and policy-makers are trying to navigate the trade-off b

6、etween facilitating testing and trials of AVs, and protecting their citizens from unproven self-driving vehicles.Transport officials cannot be expected to maintain cutting-edge knowledge on the operational safety of autonomous systems. Moreover, the industry has yet to align on a common approach to

7、demonstrating the operational safety of automated vehicles. This creates tensions between policy- makers and industry stakeholders, who may prefer that regulators trust their expertise indeveloping safe solutions, but simultaneously want to safeguard their intellectual property and other trade secre

8、ts.So far, policy-makers are largely allowing forself-certification by AV companies, while making minimal amendments to their existing road traffic regulations to enable testing of automated vehicles. A small handful of regulators are takinga hands-on approach to evaluating AVs, creating a structure

9、d, graduated testing and approval process as the basis for a licensing programme.Broadly, the Safe Drive Initiative is seeks to improve regulators decision-making abilities on automated vehicle technologies. In this paper, the World Economic Forum offers a holistic review of the leading approaches t

10、o AV policy, to identify and highlight the most effective techniques in these approaches in order to facilitate the testing and development of AVs. Following this review, we identify gaps that we believe can be addressed through a graduated, scenario-based approach developed in partnership with our

11、multistakeholder Automotive and Autonomous Mobility community.This framework will be detailed in a subsequent Safe Drive Initiative publication.In the course of this policy review, we interviewed a range of national, state and local policy-makers and advisers to understand the thinking behind the va

12、rious approaches that exist today. When coupled with a legal documentation review, this highlights a number of important success factors to developing an optimal policy environment for AVs. One message became abundantly clear: A cohesive approach to AVs is possible only when accompanied by a clear t

13、op-level vision or strategy statement for the future mobility ecosystem in the market. With a clear vision statement for the future of mobility, it becomes easy to define what role AVs can play in achieving that vision, and enabling the creation of an achievable policy roadmap for AV testing, develo

14、pment and commercialization.While this is straightforward from a strategists perspective, the reality is that many markets have various levels of state, ministerial and local regulations that may be affected byAV deployment, and these various levels of subnational regulators may also seek input on f

15、uture AV policies. This tension is all too common, and may create unnecessary hurdles for AV developers to navigate. Without a cohesive top- down strategy, it is easy to see how regulatory patchworks emerge.This project is continuing to study and propose optimal approaches to AV policy, as the techn

16、ologies move from small trials to larger pilots and deployments. The recommendations and methodologies identified in this document are intended to provide actionable insights forglobal policy-makers, while also informing our own research into AV policy.The World Economic Forum is grateful for the in

17、sights from several contributors and the analytical support of McKinsey & Company.October 2020Safe Drive Initiative:Creating safe autonomous vehicle policy1Executive summaryThere is limited global consensus on how to define milestones for AV safety, because the definition of safety is a function of

18、its operating environment.The development of automated driving is far from complete. From software to hardware, the technology is still evolving, particularly in Level 4 and 5 systems. Meanwhile, regulators are stillcatching up with this fast-growing sector, and at best are reactive in creating poli

19、cy to enable testingwith considerable variation in methodologies between nations and states. The United Nations Economic Commission for Europe (UNECE) recently proposed new autonomous driving guidance to ensure passenger safety, signalling the importance of this subject. Still the challenge of creat

20、ing policy for high-level automation remains unresolved.The industrys rapid development to this point has outpaced policy-makers response to AVs. While the technology is continuing to evolve, regulators need to understand how these development activities and future deployments will affect their citi

21、zens lives and influence their decision-making in everything from transport policy to urban development.Some countries, such as Singapore, have been highly proactive in engaging with, and providing funding for, the industry and research institutions to study the technology and create regulatory sand

22、boxes. This enables them to understand AVs intimately and shape policies that meet their own goals for mobility in their jurisdiction. Others, such as the United States, have taken a more hands- off approach, creating broad guidelines for AV development and relying on the industry to set thepace of

23、investment and deployment. In the absence of suitable federal or state regulation, cities may also feel the need to set their own guidance, creating a fragmented and varied broad policy landscape for AV companies to navigate across the country.Globally, there is limited consensus on how to define mi

24、lestones for AV safety, since the problem of defining safety invariably becomes a function ofthe operating environment. Several technically driven initiatives have proposed validation frameworks,but these have yet to be converted into workable policies. Moreover, AV developers are often reluctant to

25、 share any information about their own safety metrics regarding their operations out of fear of losing their competitive advantage, creating a siloed industry that essentially competes on safety.Some authorities, such as the California Department of Motor Vehicles (DMV), require routine data reporti

26、ng, including miles driven and disengagements, but these limited metrics are frequently refuted as poor indicators of safety by industry, academia and other governments alike. Despite this criticism, there is limited consensus among industry stakeholders as to what metrics could be used to indicate

27、system safety.While the progress of development to date has been rapid, AV technologies remain largelyunproven for many operators, and the industry has not yet been able to align on a common framework for assessing the safety of AVs necessary for full deployment. While self-certification may be the

28、preferred approach for AV providers during the testing phase, many regulators and members of the public will expect some form of independent assessment of AV safety in the future.An effective validation process must strike a balance in order to create a rigorous process that ensures safety for all r

29、oad users without demanding too much from AV providers, opening doors to costly litigation or exposing trade secrets. Policy-makers need to use this period of development to engage with the industry to advance their own understanding ofthese vehicles and establish suitable competencies to inform the

30、ir future decision-making if they want to succeed in enabling safe deployments of automated vehicles when the technology matures.Thus, the key research question behind the Safe Drive Initiative is the following: How can policy- makers ensure an AV is safe to operate in their jurisdiction?This docume

31、nt provides several contributions towards answering this question:Review of existing safety validation and regulatory initiatives, identifying best practice in existing policyCase studies highlighting a range of policy approaches, to compare policy development strategy, design decisions and methodol

32、ogiesGaps and anticipated requirements for a standardized safety validation approachIn summary, regulatory initiatives vary significantly from hands-off guidance with no formal permitting programme to structured performance evaluations. However, most policies that exist today are only temporary meas

33、ures to enable testing. The lack of commonality presents challenges for AV developers who will want to expand into different markets without going through different validation processes in every city, state and country in which theywish to operate. Moreover, a lack of the technical knowledge require

34、d to implement regulation presents a major hurdle for regulators, who seek to promote innovative technology but are unsure how to certify that it is safe for the public.An optimal AV safety assessment will benefit from a multistakeholder approach convening regulators, AV developers, safety experts a

35、nd members of the public to ensure that a system is safe to use on public roads. The exact processes will vary across jurisdictions due to geographic, political and legislative differences, but harmonizing approaches will accelerate the development and deploymentof AV technology around the world. Es

36、tablishing a standardized process will reduce uncertainty for AV providers who desire to expand into new markets, and also equip additional stakeholders such as logistics companies with the tools to compare different AV systems when choosing which to purchase.This review is the first in a series of

37、documents published by the World Economic Forum to offer holistic guidance on autonomous vehicle policy. By engaging with industry partners, regulators, academia and civil society, the Safe Drive Initiative proposes a broad and actionable framework to enable policy-makers to create a graduated serie

38、s of milestones to enable the safe deployment of automated vehicles.2Key terminologyDefining verification methodsSimulation: Highly parameterized digital twins of roads that can be used to test AV systems under many different parameter values and estimate the vehicles expected response.Test track: A

39、 controlled environment that mimics a city infrastructure and can be used to simulatereal-world driving in a lower-risk situation. Can enable verification of simulation scenarios in a real-world analogue with minimal risk.Naturalistic on-road testing: Using public roads for testing in regular traffi

40、c, typically supervised by a safety driver or remote operator. This is necessary to prove the safety of the vehicle in the real world, but also presents the most risk to the public.Other reference termsAutomated driving system: The hardware and software that are collectively capable of performing th

41、e entire dynamic driving task on a sustained basis, regardless of whether it is limited to a specific operational design domain.Autonomous vehicle (AV): A vehicle equipped with an automated driving system designed to function without a human driver as a Level 4 or 5 system under SAE J3016.Dynamic dr

42、iving task: All of the real-time operational and tactical functions required to operate a vehicle in on-road traffic, excluding the strategic functions such as trip scheduling and selection of destinations and waypoints.Operational design domain (ODD):A description of the specific operating domain(s

43、) in which an automated driving system is designed to properly operate, including butnot limited to roadway types, speed range, environmental conditions (weather, daytime/ night-time etc.) and other domain constraints.Scenario: A traffic situation within the vehicles operational design domain.Scenar

44、io-based assessment: Evaluating a system based on its performance whenexposed to a variety of predefined scenarios that correspond to its intended deployment ODD.3MotivationThis paper is intended to review current autonomous vehicle policy initiatives in order to identify the governance gaps present

45、ed by AVs. To inform this analysis, the Forum conducted an initial review of relevant AV policy initiatives worldwide. Following this, the project team conducted a series of in-depth stakeholder interviews with a range of policy-makers, from top-level national regulators to municipal-level agencies,

46、 to understand thereality and implementation of each of these policy initiatives. A series of verbatim quotes from these interviews is provided to add further depth to this document.State of the industryTo date, the industrys development efforts have focused upon three main applications of automated

47、 driving systems.Initially, the autonomous vehicle industry was heavily focused on autonomous mobility, developing vehicles for use in automated ride- hailing or transit services.Recently, there has been a pivot to developing vehicles for delivery purposes by many operators. These delivery applicati

48、ons span a range of applications, from heavy-duty trucking to small ground robots for last-mile delivery.Additionally, some original equipment manufacturers (OEMs) are developing highly automated driving systems for use in individually owned passenger cars, with a view to creating “chauffeur” functi

49、ons, for example.These applications of the technology represent very different business models, and require separate considerations in the engineering of the vehicle and its safety requirements.Delivery has become a particularly attractive application of AVs, not just because of an increased demand

50、for last-mile deliveries in light of theCOVID-19 pandemic, but because it represents a more compelling business case for a complex and expensive vehicle to be a fleet-owned asset witha potentially high degree of use. This is also true for trucking and long-haul logistics, which is also viewed as a p

51、rime application of AV technologies.As the technology development continues, autonomous vehicles are being piloted around the world to trial these solutions on public roads.Regulators are continuously challenged as they react to this evolving landscape and seek to understand the technology, and its

52、role in their future mobility systems.Current state of autonomous vehicle policy initiativesA spectrum of initiatives exist globally to study and develop AV policy solutions. From the amendment of existing traffic laws to the development of structuredlicensing programmes, the AV policy landscape rem

53、ains a fragmented one globally. Some nations and cities have been highly proactive in developing aThe AV policy landscape is currently skewed by commercial incentives.Policy expertNothing comes close to the technical knowledge companies have.City regulatorFIGURE 1roadmap for autonomous vehicles, and

54、 creating legal frameworks and standards to provide clarity to AV developers, while others have prioritized an industry- led open policy environment to facilitate innovation or devolve responsibility to regional governments.Globally, AV policy initiatives can be placed into four broad categories:Sel

55、f-certification approach: The regulator establishes a broad policy framework allowing for a range of autonomous vehicle solutions to be developed, while issuing guidelines as required on taxonomies and other definitions. The defining feature of this approach is that the government requires the AV op

56、erators to publish a self- assessed safety case, detailing the operational and organizational safety measures taken to regulators and the public alike, in exchange for a permit to test and develop on the roads.Code of practice approach: The regulator publishes guidance to AV companies seeking to con

57、duct trials on public roads with minimal amendment to existing road traffic laws. A legal code of practice establishes the expectations of the AV providers to demonstrate thesafety of their solution, along with a range of other considerations on subjects such aspublic engagement and interfacing with

58、 local authorities. There is no formal permitting as part of this process.Global overview of AV governance initiativesOperational safety assessment approach: The regulator establishes a formal assessment of the safety of the vehicles, whether on atest track, simulation or on the public road, or a co

59、mbination of the three. Passing thisassessment grants an AV company a permit to test in a given environment; the regulator may also set a graduated series of milestones.Type approval approach: A super-national body sets harmonized safety standards for vehicles across markets. This is the current app

60、roach to setting regulations for automotive safety in UNECE signatory nations. While these regulations are largely geared towards conventional vehicle safety features, UNECE is actively working on standards for automated driving to enable common performancerequirements across signatory markets. UNEC