AUTONOMOUS DRIVING - Capgemini
AUTONOMOUS DRIVINGHOW TO OVERCOME THE 5 MAINTECHNOLOGY CHALLENGES
TABLE OF CONTENTS05 . . .Introduction06 . . .Challenge 1: Assurance of systems and software- Safety- Security09 . . .Challenge 2: Sensing and connectivity- Vehicle sensors- The role of communications- The role of analytics12 . . .Challenge 3: Judgement-Decision-making- Data and change- Human interaction09 . . .Challenge 4: Architectures for managing complexity09 . . .Challenge 5: Verification and validation09 . . .Approaches to overcoming these challenges
60bnIs the Autonomous car market size forecast in 2030 vs 5.7bn in2018 with a CAGR 21.7% 17.5bnIs the full automation car market by 203018% CAGRExpected growth of the global autonomous/driverless car market,during the period 2020 - 2025220mnTotal number of connected vehicles by 2020 (vs 48mn in 2016)
INTRODUCTIONAutonomous vehicles – whether for personal transport or freight delivery – couldoffer a potentially enormous disruption to life, business and society. The possiblebenefits - reductions in accidents arising from human error, reduced cost &environmental impact of transport, liberation of time currently committed to driving,and accessibility to a wider range of users - are all theoretically addressable.Based on this context, key challenges must be overcome to achieve this: Assurance of systems and software: How can we define anddemonstrate the right level of acceptability? S ensing and Connectivity: How can we ensure the right relationshipbetween a vehicle and its environment? J udgement: How can automated systems exercise judgement? Architectures for managing complexity: How can we manage theresulting system complexity? Verification & validation: How much testing do we need, and how canwe achieve it?On this market context and this analysis, a number of implications andapproaches to overcome these challenges can be considered: The ultimate acceptability of autonomous vehicles will be a societaland political decision; consequently those involved have a duty to betransparent about their choices and the rationales for them. Assurancein the sense used by other regulated industries will in any case bedifficult to obtain. The complexity of the driving environment will demand both newsensors and new communications channels, and also increasinglysophisticated approaches to capture and interpret the information. The implementation of decision making processes must consider:-A n appropriate division of responsibility between operators,manufacturers and other parties, based on clear technicalrequirements instead of abstract goals-T he ability to correct and update decision making policies over time-T he role of human-machine interactions, will require user-centereddesign approaches to be adopted Autonomous systems will tend to high complexity, and architecturalmethods will be needed to keep costs manageable, and to make safetyassurance plausible.04 - IntroductionAltran
W hatever assurance targets are set, the complexity of vehicle’s andtheir environment will make testing challenging, so:-T est approaches capable of supporting massive and wellcharacterised test programmes are needed.-E vidence gathered from a wide range of assurance methods (notonly dynamic testing) will need to be used. In addition, this technological domain is changing rapidly;companies – and governments – will need to invest to track emergingtechnology trends.CONNECTIVITYHow can we ensure theconnectivity with itsenvironment?JUDGEMENTHow can automatedsystems exercisejudgement?VERIFICATION &VALIDATIONHow to ensurereliability?ASSURANCEHow can we definethe right level ofacceptability?COMPLEXITYHow to manage therequired systems'complexityAltranIntroduction - 05
.Motivation for move to autonomous vehicleLooking at next generation of cars, we notice a wide range ofmotivations for a move to autonomous vehicles, and a potentiallyenormous disruption to life, business and society. Possible benefitsare all theoretically addressable, and initial demonstrations andexperiments (from traditional OEMs and from technology companies)are encouraging.A reduction in road traffic casualtiesAccording to World Health Organization, a specialized agency of theUnited Nations, the number of road traffic deaths worldwide hit 1.35million every year not including injured or disabled people. Latest studiesshow human error to account more than 90% to road fatalities leaving highimprovement opportunities for autonomous driving technologies. Inparticular, more than half road fatalities occur among pedestrians, cyclistsand motorcyclists. Controlling for variables such as fatalities caused by caraccidents could further show the potential of autonomous driving features.Although there is a lack of consistent global estimates, the WHO estimatesthat the cost of injuries is approximately 3% of a typical country’s grossnational product. [1]Reduction in social & environmental costs of drivingWith as many as 9 billion people are predicted to live in urban areaswithin the next 25 years, automakers are under pressure to reduce theenvironmental and social impact of driving. The adoption of autonomousfeatures in cars will lead to environmental benefits: autonomoustechnologies have the potential to easing traffic flow by allowing optimizedacceleration and deceleration, thus reducing fuel consumption andemissions, and to allow better arbitration of roads & parking to reducetheir impact [2]. Changes in vehicle use may also bring benefit: autonomycan facilitate vehicle sharing, and for each car-sharing vehicle on the06 - IntroductionAltran
streets more than 20 vehicle sales are forecast to be deferred. For higherspeed journeys, benefit could ultimately be derived from technologiesallow vehicles to follow each other closely (platooning), reducingaerodynamic drag by 20-60%.Economic benefits of making travel time productiveIn its blue paper about Autonomous cars Morgan Stanley states driverlesscar could contribute 1.3 trillion in annual savings to the US economyalone and 5.6 trillion in global advantages. Focusing on productivity, thepaper further suggests gains would come to 507 billion annually in theUS. Such benefits accrue to consumers who experience a transformation inthe ease at which they can travel, which in turn generates wider economicbenefits. [3]Potential improvements of access to mobilityWhile driverless technologies are being implemented first in luxurysegments, once fully autonomous cars are available, significantimprovements will be held in access to mobility. Such technologies willin fact act as key enablers for people with physical limitations, the young,and the (increasingly numerous) elderly. A UK study shows that about1,45 million people are facing mobility issues and that is only taking intoaccount over 65 years old and in England alone [4].But what challenges must be overcome to achieve this vision? Isour technology, and the industries which support it, able to achievethese benefits?[1] www.who.int/violence injury prevention/road safety status/2015/en/[2] vironmental-impact/, www.rand.org/content/dam/rand/pubs/research briefs/RB9700/RB9755/RAND RB9755.pdf, r Cebr-Cost-of-Congestion.pdf[3] -future-is-now, d2e263968d9[4] ise-old-age.htmlAltranIntroduction - 07
CHALLENGE 1ASSURANCE OF SYSTEMSAND SOFTWAREHow can we define and demonstrate the right level of acceptability?Assurance (n): a positive declaration intended to give confidenceoxforddictionaries.comIn order to product or operate an autonomous vehicle, we mustprovide a range of stakeholders with assurance that the vehicle willoperate safely. This is no different from the principles which apply tomanually-controlled vehicles or to any other system we deploy. But, canwe construct and maintain systems (across a potentially large vehiclepopulation) that give us necessary confidence in their operation? Whatcriteria will determine the acceptability of autonomous operation? Howcan confidence be maintained in the face of malicious activity?.SafetyThe process of providing this confidence shares many factors withexisting systems and vehicles:The difficulty of bounding responsibilitySafety applies to a road system, not a car; safety cannot be measureddirectly, only judged from examination of dynamic interactions betweencomponents and effects outside the system boundary.The difficulty of characterising the environmentThere are features of day-to-day driving which will be difficult tocharacterise for development or to replicate for testing: temporaryinfringement of traffic laws, snow on road markings, hand signals frompolice officers at an accident and other everyday “black swan” situations.But there are also factors specific to autonomous road vehicles:Automotive transport is much less regulated (and quantitatively lesssafe) than other environments such as rail or aviation; the road systemis also already prone to single-point failures (that is, misbehaviour of asingle vehicle or pedestrian).08 - Assurance of systems and softwareAltran
Autonomous vehicles will make mistakes that are different fromthose that humans make because they sense the environmentdifferently – this has implications both for the vehicle itself (as thedesign must not simply seek to replicate human behaviour) and forother road users (whose safety may be jeopardised by the presenceof entities that don’t respond as expected). Functions that replace the driver in certain situations, but whichmust be replaced by the driver in situations they cannot handle,raise the question of why the (uninvolved) driver will be effectiveonce the automation becomes ineffective. Automation will reducedriver attention to hazards. (Control cannot be returned to the driverinstantaneously, unless there is “look-ahead” prediction that detectsa difficult situation and can alert the supervising human in goodtime, without triggering a panic (over)reaction).The current regulatory framework for road vehicles, exemplified by theUNECE Transport Regulations and ISO 26262:2011, is not intended toaddress such issues and is likely to need substantial evolution in orderto do so. There is a risk that shifts of some responsibility from the driverof a manually driven vehicle to the manufacturer of an autonomousvehicle – which may well need legal and administrative changes – willtrigger an over-reaction, resulting in setting impossibly high standardscompared to human drivers.AltranAssurance of systems and software - 09
.SecuritySecurity is a particular concern with computer-based systems, and itunderlies any other aspect of assurance, because if a system is open tomalicious modification, no other behaviour can be relied upon.Attention will be paid to autonomous vehicles, both by potentialattackers and by those attempting to maintain security, because thepotential impact of a risk – perhaps even multiple simultaneous failuresacross a whole vehicle fleet – could be so great. The likelihood thatautonomous vehicles will be networked presents two further aspects:connection to off-board computer facilities (or the cloud) opens newvectors of attack, but also enables cloud based behavioural monitoringof the vehicle fleet which can identify malicious activities early.Security is an emergent property of a system in achanging environment and we believe this can only beaddressed by a combination of approaches.Security is a particular concern with computer-based systems, and itunderlies any other aspect of assurance, because if a system is open tomalicious modification, no other behaviour can be relied upon.Attention will be paid to autonomous vehicles, both by potential attackersand by those attempting to maintain security, because the potential impactof a risk – perhaps even multiple simultaneous failures across a wholevehicle fleet – could be so great. The likelihood that autonomous vehicleswill be networked presents two further aspects: connection to off-boardcomputer facilities (or the cloud) opens new vectors of attack, but alsoenables cloud based behavioural monitoring of the vehicle fleet which canidentify malicious activities early.A set of principles we have found useful elsewhere [5] is:- Know Your EnemiesUnderstand the security risks posed by a system and form acomprehensive policy to deal with them.- Take Security to the EdgeAddress security from end devices to central services, and frominitialisation to disposal.10 - Assurance of systems and softwareAltran
- Know What You’re Talking ToUnderstand the identities, roles and authorisations of people andequipment. Address the provisioning of new identities, maintenance,change of ownership, and withdrawal of trust.- Create A Strong NetworkEnsure communications are resilient and resistant to attack.- Don’t Trust It, Watch ItMonitor behaviour for signs of attack, don’t rely on fixed defences. UseSIEM (Security Information and Event Management) techniques includingadvanced analytics.- Build It RightMinimise the vulnerabilities exposed to an attacker. Use security-orientedarchitecture, separation of security domains, highly-assured softwareand hardware components, and generate assurance evidence duringdevelopment.- Base On Firm FoundationsUse trustworthy services for communications, computation, storage andmanagement.[5] Seven principles for achieving security and privacy in a world of Machine-Driven Big Data,Altran White Paper 2016, /PDF/AltranPosition paper WEB V2PDF.pdfAltranAssurance of systems and software - 11
CHALLENGE 2SENSING AND CONNECTIVITYHow can we ensure the right relationship between a vehicle and its environment?How can autonomous vehicles gain sufficient information on theirenvironment to operate efficiently and safely under all circumstances?What sensors, and what analytics applied to sensor data, will berequired? What communications channels – vehicle to infrastructure(V2I) and vehicle to vehicle (V2V) – will be used? How can external databe used? How do we manage transient connectivity?.Vehicle sensorsThe process of providing this confidence shares many factors withexisting systems and vehicles:The difficulty of bounding responsibilitySafety applies to a road system, not a car; safety cannot be measureddirectly, only judged from examination of dynamic interactions betweencomponents and effects outside the system boundary.The difficulty of characterising the environmentThere are features of day-to-day driving which will be difficult tocharacterise for development or to replicate for testing: temporaryinfringement of traffic laws, snow on road markings, hand signals frompolice officers at an accident and other everyday “black swan” situations.But there are also factors specific to autonomous road vehicles:Automotive transport is much less regulated (and quantitatively lesssafe) than other environments such as rail or aviation; the road system12 - Sensing and connectivityAltran
.The role of communicationDirect sensing represents the simplest, but not the only, means for avehicle to build a model of its environment, it may also directly receiveenvironmental information via (wireless) communications channels.These channels may be established between vehicles (vehicle tovehicle, v2v) or between a vehicle and fixed infrastructure (vehicle-toinfrastructure, v2i). Possible applications include:- Signalling of presence and planned behaviour between vehicles(radio brake lights)- Sharing of environmental information among local users via v2vcommunication (eg stationary traffic or icing warnings)- Warning and control information distributed by v2i channels (radiotraffic lights, road signs)The major challenge regarding such systems is the level of dependencewhich can be placed on communications systems in implementingsafety-related functions. Although v2x technology has been developedand standard promulgated, there are limits to the assurance thatcan be established for radio communications, particularly with (orbetween) rapidly moving vehicles, limiting their applicability forsafety-critical functionality. Nevertheless, if the travel efficiency benefitsof autonomous vehicles (or even of advanced driver informationsystems) are to be realised, a level of information must be shared inreal-time, although whether this is through automotive-specific v2v orv2i technologies, or simply over standard mobile (3/4/5G) networks isopen to question.AltranSensing and connectivity - 13
.The role of analyticsSuccessful designs will make the greatest possible use of the dataavailable from their sensor suites; signal processing and analytics insupport of sensor interpretation will be key technologies.Examples include:- Sensor fusion to take advantage of multiple input sources- Vision processing for the extraction of road features and signage- Object recognition, and even intent recognition [6], to facilitateaccident avoidance and to improve trajectory and maneuverplanning and execution.The techniques – such as machine learning – used to achieve theseresults are computationally intensive and difficult to verify by traditionalmeans. To bring such systems into mass production will requireadvances in both implementation and verification.[6] See www.mrt.kit.edu/mitarbeiter 3269.php
CHALLENGE 3JUDGMENTHow can automated systems exercise judgement?How can autonomous vehicles be constructed to manage the (oftenconflicting) expectations placed on them? Can algorithms make thesubjective and ethical decisions required of human drivers? How canexternally defined policies be communicated, validated, and updated?How will humans (inside and outside a vehicle) interact with it? Howmust user experiences change to adapt to autonomy?.Decision-makingA significant amount of discussion [7] has been published about theapparent need for autonomous vehicles to make “ethical” judgementsabout the consequences of particular actions, even extending to surveys [8]of public attitudes.Autonomous function certainly changes some aspects of responsibilityand liability compared to manual driving – actions such as choosingan appropriate speed for prevailing conditions, which are the soleresponsibility of a human driver in a manually driven vehicle becomebehaviours of a product which has a manufacturer, a designer, and avendor as well as an operator. The legal and commercial aspects of thischange are beyond the scope of this paper, but the expectations raisedabout decision-making functionality cannot be ignored.We can argue that this discussion is of little practical relevance becausethe decisions taken during the design of autonomous vehicles are notexpressed at a level where human interpretations are possible. Many otherproducts with potentially lethal consequences are regularly used withoutsuch concerns being raised, nor are questions of moral philosophy oftenincluded in driving tests.[7] For example, www.people.virginia.edu/ njg2q/ethics.pdf and www.driverless-future.com/?page id 774#ethical-judgements.[8] The social dilemma of autonomous vehicles, Jean-François Bonnefon, Azim Shariff, IyadRahwan3, Science 24 Jun 2016: Vol. 352, Issue 6293, pp. 1573-1576AltranJudgment - 15
The challenge of defining and quantifying autonomous vehicle behaviourmay stem from the variability and complexity of the situations in whichdecision making will be required and the numerous and frequentexposure of people to the consequences of those decisions that willarise if autonomous vehicles are widely deployed. (Tesla already (May2016) report 780m miles driven in vehicles equipped with their autopilothardware, and 100m miles driven with autopilot active [9].)In complex and ill-characterised road environments, both the algorithmsadopted and the measures used to assess them will be statistical, ratherthan entirely absolute, in nature. This contrasts with other domains, where,for example, railway control systems may be protected by interlockingsystems with binary (on/off) specifications and implementations, oraviation, where substantial effort is spent in verifying control systemsagainst precise abstract specifications, and the operational environment isrigorously controlled by highly-trained and monitored staff (both pilots andair traffic controllers). The implications of the consequential growth in bothtest demands and available test data are considered further below.Implementation technologies for autonomous vehicles are focussed moreon empirical observation of driving environments and decisions usingtechniques such as machine learning, where machine states and actionsare characterised and models are trained by assigning rewards or costs forthe system being in certain states.While individual parts of such systems can be expressed and validatedin absolute terms – we can specify and test that a vehicle never drivesthrough an obstacle that is adequately represented in its situationalmodel – absolute tests of overall system behaviour that tie to real-worldobservations are unlikely to be feasible. Overall system behaviour is morelikely to be able to be validated statistically, with tools such as confusionmatrices or ROC curves. [10][9] ata/[10] Receiver operating characteristic16 - JudgmentAltran
.Data and changeBecause they operate in a rich and changing environment, autonomousvehicles are likely to need configuration and reference data which isliable to change. The quality and maintenance of quality, of such datais crucial, as autonomous vehicles will be much more dependent onavailable data than manually-driven ones. Some data (for example, trafficmanagement policies in specific jurisdictions, or system configurationdata) may be relatively limited in volume and be amenable to rigorouschange control and regression test processes. Others, such as map and‘electronic horizon’ data (if used) will be of such volume and complexitythat comprehensive testing will be difficult, and will be captured byprocesses that lack the stringent independent checks that are used fordata preparation in other industries.A defence-in-depth strategy, with mechanisms in place to detect potentialerrors in data by cross checking with observation, would seem sensible.Human interactionAutonomous systems will radically change human interactions withvehicles, both for their users and potentially for third parties. Much ofthe interface functionality may be relatively standard (setting objectives,querying status and progress) but a crucial new interaction will arise whenthe autonomous system needs to pass control back to a human operator:the question of whether this can be done at all, in an acceptably safemanner, is still controversial.Particular concerns include:- The time delay necessary to alert the driver from an ‘eyes-off’condition, and whether autonomous function can maintain vehiclesafety for such a period.-H ow control can be transferred in a way which avoids sudden overreaction or panic on the part of the driver.A user-centred approach will be necessary to address such issues.AltranJudgment - 17
CHALLENGE 4ARCHITECTURE FOR MANAGINGCOMPLEXITYHow can we manage the resulting system complexity?Numerous interacting functions controlled by distinct stakeholders mustcome together to achieve autonomous driving. Can the architectures ofour systems manage the consequential level of complexity? How willthe industry adapt to the era of the Software Defined Vehicle?The autonomous control of a vehicle implies a large number ofcooperating functions. At a high level these functions include:1. An interface to an end-user who needs to provide goals to the vehicleincluding destination, preferred route characteristics, intermediate stopsand possibly a target arrival time2. A navigation system capable of planning the appropriate route,determining position (against a map), developing machine executableinstructions to follow the route based on current position in real-time3. An environmental perception system which determines the externalsituation and in particular threats and safety related constraints (othervehicles, objects, pedestrians etc.)4. A vehicle context system which maintains a model of the vehicle stateincluding speed, fuel levels or health status5. Active safety system capable of using the available data from theenvironment and the vehicle context to plan manoeuvres and ensuresafe actions by the vehicle6. A vehicle control system which can take instructions from the navigationsystem (run left/right etc ), the vehicle and environment context and,with the permission of the active safety system, issue control commandsto the vehicle actuators7. Actuators which perform actions on the physical driving components(steering, powertrain, braking, etc )8. Sensors which provide the data to the various systems18 - Architecture for managing complexityAltran
The entity which results from the integration of these functions will exhibithigh levels of function intelligence, a large number of interaction pathsand a very large potential state space and will operate in a dynamic andevolving environment.In taking the above in consideration it should be evident that creatingsuch a system presents a major challenge of dynamic complexity.Such complexity leads to several serious issues, notably the cost andtime of integration, the challenges of testing, assurance and dataquality (discussed elsewhere) and the difficulties of ensuring adequateperformance and confidence throughout the life of a product which mustbe maintained and adapted over many years.Traditional automotive architectures are based on considerations offunctional domain, supply chain structure and historical networkingheritage. In reflecting this, a current vehicle is structured as a collectionof Electronic Control Units (ECUs) wired together by CAN. Each ECUgenerally covers a unique domain and comes from one supplier.Interactions between the ECUs are through the CAN telegram protocols.This approach is struggling with the rise in complexity represented by themove towards advanced ADAS and autonomy.The fundamental driver in this evolution is increase in software volume( 100M LOC and increasing) and the complexity involved in ensuring(and proving) that such a volume of software provides its functionality withappropriate performance and in the presence of constraints on power,BOM, supply chain and timescales.In the face of these issues the traditional architecture suffers from thefollowing shortcomings:- Ad Hoc and diverse approaches to such things as lifecycle, errormanagement, thread priority management and communications- BOM inefficient approaches to peak CPU requirements- Ad Hoc and divergent data and software interaction semantics- Low bandwidth networking- Timing divergence- Supplier lock-in and lack of modularity at a software level- I nflexibility and extreme brittleness in system re-configuration and reengineeringAddressing these problems requires a new architectural approach with thegoal of satisfying the traditional needs of performance and BOM controlwith the direct attack on the problems raised by complexity.AltranArchitecture for managing complexity - 19
Future architectures are likely to be based on a number of key strategies:- A rethink of the HW structure towards a more centralised computeresource with more power available to be managed for peaks acrossfunctions. This approach also offers a significant reduction in weight andcabling complexity in the vehicle. The Domain Controller approach is anexample of such a strategy.- The introduction of mainstream IP based networking, albeit with somespecific extensions for determinacy where required. This will serve tosupport mainstream development styles as well as adding bandwidth.- The introduction of a flat data plane for all data access, including forVideo and Audio. This will enable data to be acquired by software in anear arbitrary manner and with less configuration issues.- A stricter definition of citizenship for software elements. This will implythat certain aspects of software behaviour will be uniquely defined at
syntax and semantics for all software. One example would be softwarelifecycle where a specific and unique API would be enforced for allsoftware elements.- A support enforcement of location transparency in all softwaretransactions. This means that all software will interface to an underlyingand transparent communication layer. This implies that software will notneed to know the system topology for communication.-D irect architecture support for security structures and technology. Thiswill allow security to be managed as a configuration activity during systemintegration and based on standard system elements.-D irect architecture support for Safety measures. This will allowsafety critical aspects of the system to be isolated and monitored in astandardised manner.The global change this represents for the architecture is that the vehiclehardware and the base software will present a unified feature platformor the whole vehicle and not just for an individual ECU. This changewill support a more disciplined approach to system engineering whileproviding key approaches to tackling the complexities of the system.AltranArchitecture for managing complexity - 21
CHALLENGE 5VERIFICATION & VALIDATIONHow much testing do we need, and how can we achieve it?How can autonomous systems be tested to the levels of confidencerequired? What data sources, and what reference cases, will berequired? How much testing will be needed? How can
improvement opportunities for autonomous driving technologies. In particular, more than half road fatalities occur among pedestrians, cyclists and motorcyclists. Controlling for variables such as fatalities caused by car accidents could further show the potential of autonomous driving features.
Dr. Brian Whitmore - Capgemini VR & AR meets Microservices Ben Scowen Capgemini Gary Baptist Capgemini Unblocking innovation and agility Roman De Oliveira Mulesoft 15:15 - 15:45 Driving transformation through insight Lee Brown Capgemini 15:45 - 16:15 16:30 - 17:00 SAS Viya: Rapid innovation in an open, connected world Paul Jones SAS Blockchain .
Page 2 Autonomous Systems Working Group Charter n Autonomous systems are here today.How do we envision autonomous systems of the future? n Our purpose is to explore the 'what ifs' of future autonomous systems'. - 10 years from now what are the emerging applications / autonomous platform of interest? - What are common needs/requirements across different autonomous
autonomous driving solutions is a highly valuable skill in today's software engineering field. Robot Operating System (ROS) is a meta-operating system that simplifies the process of robotics programming. This master's thesis aims to demonstrate how ROS could be used to develop autonomous driving software by analysing autonomous driving
SAP and Capgemini are working together as part of this initiative to address the technological and organizational challenges of their clients going digital. Capgemini Digital Control Room Analytics leverages Capgemini’s proven Digital Transformation Framewo
cloud, ata, AI, connectivity, software, igital enineerin and latfors. The Group reported in 2020 global revenues of 16 billion. Get The Future You Want Visit us at www.capgemini.com MACS 3691_10-2021 For more details contact: Valérie Perhirin Managing Director, Insight Driven Enterprise - Capgemini Invent valerie.perhirin@capgemini.com
Capgemini had deployed Cisco ACE load balancers to load balance web, mail and other applications in the multi-tenant data center. "Capgemini shares the same hardware for a lot of different customers," says Sven Verbeek, Capgemini's technical consultant. A major requirement for new networking hardware is suitability for multi-tenancy; it
Audi’s initiative &Audi aims to contribute to the introduction of autonomous driving in the inter-ests of individuals and of society. The initiative’s approach is to tackle the challenges along the way to autonomous driving together with various social stakeholders. Since 2015, Audi ha
Article 505. Class I, Zone 0, 1, and 2 Locations Figure 500–2. Mike Holt Enterprises, Inc. www.MikeHolt.com 888.NEC.CODE (632.2633) 25 Hazardous (Classified) Locations 500.4 500.4 General (A) Classification Documentation. All hazardous (classified) locations must be properly documented. The documentation must be available to those who are authorized to design, install, inspect .
