Publicada el 20 de Noviembre de 2020

The future is here. Autonomous vehicles are progressively arriving to our lives, and in the years to come, they will become a key part of our cities. Nowadays conventional vehicles share our urban spaces with different types of new mobility solutions like electric bicycles, scooters, or carsharing among many other options. Autonomous vehicles are not less, and we are experiencing how they are being integrated in our smart cities, like the self-driving taxis services in Shanghai, or self-driving buses in Europe. But, what do we need to optimize their functions and give a valuable service to users? They need a well-developed digital infrastructure.  

Reason number 1: Non-Line of Sight (NLOS) and redundancy

I will not elaborate much on this one. Although the Federal Communications Commission (FCC) does not seem to share this view, as made obvious by the Notice of Proposed Rulemaking FCC-19-129A1, NLOS scenarios underlay many of the use cases advocated by those supporting Cellular Vehicle to Everything (C-V2X). Rel 16 NR 3rd Generation Partnership Project C-V2X has proposed a myriad of advanced use cases which together with the 3rd Generation Partnership Project Rel 15/16 Basic Safety Message speak by themselves.

Reason number 2: Supervising the reliability of Autonomous Vehicles during the whole lifecycle

Autonomous vehicles need to be able to be operated correctly (that is, to be reliable) during a period which may extend up to twenty years. The multidimensional constraints of electronics, software and environment to optimize the Size, Weight, Power and Cost (SWaP-C) of the automotive industry, combined with human behavior, challenge the expectations of having the AV operating as expected for such a long period in different manners.

  1. Electronics pose an interesting set of challenges to address. The first problem comes with the need to increase the available power to support the increasing demand for Artificial Intelligence/Machine Learning. Although uncharted territory, this leads to increasing transistor densities to 5nm chips. Granted that automotive functional safety standardization (International Standard Organization (ISO) 26262 or ISO/PASS 21448) and in field period testing (either periodic, or every time the car the user hits the ignition or stops at a light) should help to ensure functional safety, even more if (UL4600- Standard for Safety for the Evaluation of Autonomous Products) if this is based on the assumption that the vehicle will have no responsible human driver. Nevertheless, maintaining this status during the whole lifecycle of the vehicle at a moment where the number of chips in every car is ballooning is not a trivial task.
    On top of this, another added problem comes when considering that ISO 26262 is founded on a culture of cooperation among suppliers where post-mortem diagnosis is disclosed to the supply chain in order to perform an appropriate root cause analysis. This is no easy task when one considers that the traditional supply chain, normally structured in different tiers, is being shaken up by Integrated circuit (IC) designers and manufacturers trying to get up the ladder to compete with Tier-1, and Original Equipment Manufacturer -OEM- (newcomers and incumbent alike) are willing to design their own complete systems contacting the lower tiers directly. Industry wide reliability along the whole lifecycle of the autonomous vehicle is a long shot.
  2. Software and Over the Air Updates (OTA) present another thrilling challenge.
    Agile has brought a large transformation to an industry still largely used to waterfall software engineering. Nevertheless, scale Agile developments in safety critical industries is complex, and although when paired with OTA reduces the time to market, it also increases the chance of malfunctions.
  3. Finally, there is the issue of the environment and human behavior.
    Although a less fancy topic, the owner’s behavior when cleaning and maintaining car sensors is another topic deserving attention. Many of the OEM place Advanced Driver Assistant Systems (ADAS) sensors between the windshield and the wiper path, and the radar below the bumper. This would solve part of the problem, along with the automatic fluid jets we see in other sensors which are more exposed. Nevertheless, the huge number of sensors needed for levels 4 and 5 will increase the challenge to avoid exposed areas, and therefore guarantee that they will keep working as expected in all circumstances including rain, dirt, snow, fog, moisture in the camera lenses… and eventually minor car accidents.

This is where technology supervision comes, as it seems reasonable to assume that in the conditions presented above, the common public good and road safety would be well served by a function (integrated into the overall digital infrastructure supporting the deployment and operation of autonomous vehicles) independent of the vehicle, and able to supervise the on-going functional safety of autonomous vehicles travelling our roads. This should identify potential risk scenarios and supervise the dynamic behavior of every autonomous vehicle in them (Intel’ s Responsibility-Sensitivity Safety (RSS) may be a good foundation). It is difficult to imagine a case where all this technology tightly packed into each autonomous vehicle (and subjected to aging, interoperability issues, OTA malfunctions…), and which shares lanes, ramps, interchanges… with all sorts of vehicles, connected and non-connected, does perform without any level of third-party supervision and control for the whole lifecycle of the autonomous vehicle, even when considering Operational Design Domain (ODD) geofenced areas. This layer should probably grant traffic officers a limited capacity to eventually enforce a vehicle when it is deemed to be a potential danger. Is this the same service provided by the teleoperation control centers? Even when those advocating a mandated teleoperation may be right, this is not what I am discussing here, because this is a universal, statistic and independent service operated by the public entities in charge of road safety. Its role would not be to assist (teleoperate) but to detect. Does it cover the whole US road network of over 4 million miles? Not likely, this seems to be deployed on an as-needed basis. Is it deployed in the same manner, technology and density regardless of the environment (urban versus interurban)? Not likely.

Reason number 3: Digital infrastructure will facilitate the dynamic management of our physical infrastructure

The view that the author of the article “deconstruct the myth of AV and 5G” was proposing is very narrow and neglects the role of the traffic engineer. It conceives the autonomous vehicles as an entity in the immediate vacuum, isolated from its environment, and where the value of digital infrastructure is limited to contribute to the situational awareness of the vehicle which is (apparently) already guaranteed by the on-board sensors and Artificial Intelligence/Machine Learning applications.

The future is here

No one (at least not me) discusses the need to increase spending to fix our crumbling infrastructure and meet an increasing demand, but this will not be done overnight. In the meantime, to increase the current capacity of our roads, promising results have been achieved by integrating (for example) variable speed limits control with ramp metering and route guidance, or dynamic tolling with reversible lanes. Nevertheless, they are all depending on using Microwave Vehicle Detection System (MVDS) to sense traffic volumes, and Dynamic Message Sign (DMS) to address vehicles, as they lack the capacity to sense and address every vehicle in the traffic flow as an independent entity. Cellular Vehicle to Everything  (C-V2X) capacity to use the downlink to influence (or direct) the behavior of every vehicle (autonomous or not) by multicasting is a game changer, as it will allow it to influence every vehicle (being driven  by a human or not) as a separate entity, opening the door to fine-tuned the demand to the capacity of each lane, all of this without considering the auction off road segments or use of road over a time period as it was proposed a few weeks ago in this article in Forbes. In summary, Digital infrastructure will transform the way we manage infrastructure. From a mostly static approach where every component (lanes, curbs, ramps…) is assigned based on its capacity, to a dynamic assignment in terms of Willingness to pay (WtP), demand, time, weather, incidents and or multifunction. This will add to the significant amount of variable data (traffic lights, detours, accidents, incidents) result of a myriad of changing road conditions.

Conclusion

Research and Development (R&D) in the Autonomous Vehicles industry is relentless and tsunamic in volume but keeping our streets and roads safe is a priority while we optimize the usage of the public space. The deployment of autonomous vehicles needs to be supported by an efficient digital infrastructure, which will play a critical role during a transition period in which vehicles with all sorts of communication capabilities and automation levels will share our roads.

Acronyms

3GPP: 3rd Generation Partnership Project

ADAS: Advanced Driver Assistant Systems

AI/ML: Artificial Intelligence/Machine Learning

AV: Autonomous Vehicle

BSM: Basic Safety Message

C-V2X: Cellular Vehicle to Everything

DMS: Dynamic Message Sign

FCC: Federal Communications Commission

GNSS/IMU: Global Navigation Satellite System/Inertial Measurement Unit

HD: High Definition

ISO: International Standard Organization

MVDS: Microwave Vehicle Detection System

NLOS: Non-Line of Sight

NPRM: Notice of Proposed Rulemaking

ODD: Operational Design Domain

OEM: Original Equipment Manufacturer

OTA: Over the Air Updates

RSS: Responsibility-Sensitivity Safety

SCMS: Security Credential Management System

SPA2: Scalable Product Architecture

SPaT: Signal Phase and Time

SWaP-C: Size, Weight, Power and Cost

TTS: Traffic Technology Services

UL: Underwriter Laboratories

UMTRI: University of Michigan Transportation Research Institute

VOD: Visual Odometry

WtP: Willingness to pay

Written by Julià Monsó the 20 de Noviembre de 2020 con las etiquetas: Artificial intelligence Big Data Infrastructures Innovation Self-driving car Transport systems

No comments, yet

Inicia sesión

Para guardar tus lecturas y seguir en otro momento, necesitamos saber quién eres

¿Has olvidado tu contraseña?

Not registered yet?

You can also login with:

Sign up

Enter your email address and we will send you an email to activate your profile

You can also login with:

¿Has olvidado tu contraseña?

Introduce la dirección de correo electrónico con la que te registraste para recuperarla.

¿Has olvidado tu contraseña?

Password changed

Aviso

No se ha podido cambiar su contraseña de acceso.

¿Has olvidado tu contraseña?

Please, check your email to get the confirmation link

Aviso

No hay ningún usuario registrado con esa dirección de correo electrónico.

Aviso

Este usuario no tiene permitido el restablecimiento de su contraseña.

Sign up

Check your email

Please, click on this link to get advantages of having a user account

Aviso

Ya estabas registrado con este correo electrónico

Aviso

Sorry we have had a problem completing your registration, please try again. .

Aviso

Lo sentimos, pero ese código de validación ya se ha usado en el registro de una cuenta de usuario.

Complete your registration info

¿Qué te interesa?

Selecciona los temas que te interesan y te enviaremos el contenido relacionado.

How often would you like to receive updates?

Newsletter