Vehicle Autonomy Is an Automation Challenge

Overview

Autonomous cars, trucks, and aircraft are creating new ways to move people and deliver packages. The widespread adoption of autonomy will profoundly change transportation.  The race to certify these land and air vehicles will give the winners a distinct advantage in establishing a foothold in these new service-oriented transportation markets.  These markets will impact the jobs of those who drive or pilot vehicles and alter ownership and service business models.

Numerous new technologies are competing and evolving to support autonomous operation; many funded by organizations fighting for a position in the new world of autonomous vehicles.

Some of the key technologies in the vehicle autonomy market area for both land and air applications are lightweight and lower cost lithium batteries, improved lightweight electric drive motors, a range of sensor technologies, computers using artificial intelligence and machine learning algorithms to safely achieve autonomy, and communication systems that allow coordinated operation.  For land vehicles, there are competing technologies using sensors for LIDAR, cameras, radar, ultrasonic, GPS, and microphones as the inputs to powerful processors that will provide situational awareness and decide vehicle actions for steering, acceleration, and braking.  Drone autonomy is developing with a focus on ADS-B, a GPS based technology that is a key element of the evolving NASA UTM (Unmanned Traffic Management) system.

Levels of Vehicle Autonomy

For land vehicles, SAE J3016 describes the different levels of autonomy:  

• Level 0: No automation
• Level 1: Driver assistance - adaptive cruise control
• Level 2: Partial automation - two or more advanced driver assistance systems (ADAS)
• Level 3: Conditional automation - full control and operating during select parts of a journey
• Level 4: High automation - capable of completing an entire journey without driver intervention, even operating without a driver at all
• Level 5: Full automation complete hands-off, driverless operation under all circumstances

Trials are currently under way for Level 4 vehicles, but so far, no car or truck has been approved to freely operate at Level 4 or Level 5 autonomy.

Safety Certification of Vehicle Autonomy

In the past, vehicle drivers and aircraft pilots have been certified with specific licenses that define the vehicles or aircraft they can operate.  With vehicle autonomy we still need to certify.  But now, that certification relates to the automation of the vehicle for its intended use.  The ability to operate safely without human intervention needs to be rigorously designed, built, verified, and validated to achieve vehicle certification.

For industrial automation, the International Electrotechnical Commission's (IEC) standard 61508 defines the safety integrity level (SIL) using requirements grouped into two broad categories: hardware safety integrity and systematic safety integrity.  A device or system must meet the requirements for both categories to achieve a given SIL.  SIL level is a measure of the risk of a dangerous failure.  SIL 1 is the lowest requirement and SIL level 4 would have a much greater risk reduction factor.

In a similar structure, vehicle autonomy uses an ASIL standard ISO 26262. The ASIL, or automotive safety integrity level, is established by performing a risk analysis of a potential hazard by looking at the severity, exposure, and controllability of the vehicle operating scenario.  The safety goal for that hazard in turn carries the ASIL requirements.  The standard identifies four levels: ASIL A, ASIL B, ASIL C, and ASIL D.  ASIL D dictates the highest integrity requirements on the product; ASIL A the lowest.

Individual systems in a car (airbags, power steering, sensors, etc.) are rated with the ASIL methodology; but overall functional safety of an autonomous vehicle must be rated on a systems basis.

Similar risk-based safety standards exist in the aviation and rail industries.  The table below shows the standards and the approximate mappings across the domains.

In July 2019, a coalition of eleven companies — Aptiv, Audi, Baidu, BMW, Continental, Daimler, Fiat Chrysler Automobiles, Here, Infineon, Intel, and Volkswagen — published a whitepaper: “Safety First For Automated Driving” (SaFAD).   This paper describes a framework for developing, testing, and validating “safe” autonomous vehicles.  This document was produced to fill in the gaps of ISO 26262 and help state, federal and other international agencies develop appropriate rules and regulations.  The document describes twelve guiding principles and has received much praise and attention as a major step that will clarify the development and certification process.

The processes and procedures described in the SaFAD paper parallel the similar industrial procedures used to verify and validate control systems in the pharmaceutical and nuclear industries.  Vehicle autonomy and industrial automation share many common design structures.  Control and safety depend on measurement devices, logic and algorithms that monitor these sensors, and outputs that act to keep the operation safe.  Both industrial and vehicle situations require accurate, well-calibrated sensors and need to ensure safety even when sensors fail.  However, unlike most industrial automation, the logic and control algorithms for vehicles will be using non-deterministic neural network processing of sensor information based on training from massive data sets.  Industrial control systems can be tested with high-fidelity process models, but that option is not available for the enormously large number of situations that vehicles will need to deal with.  As described in the SaFAD document, validation testing must be based on statistics gathered by operating in the actual environment.

It is not hard to imagine scenarios in which passenger car autonomy could cause crashes and fatalities.  On the other hand, crash data on our roads today suggests that 95 to 98 percent of car fatalities are caused by human error. With validation testing of autonomous vehicles on real roads, current progress indicates that autonomous safety will likely exceed human safety and automation will likely save lives, reduce CO2 emissions, and reduce insurance premiums.

In a similar line of reasoning, speaking at the Uber Elevate conference, John Langford, president and chief executive officer of Aurora Flight Sciences, claimed: "Certifiable autonomy is going to make quiet, clean, and safe urban air mobility possible."  

Industrial Automation Parallels with Vehicle Autonomy

Anyone who has designed tested and lived with process controls or safety systems that operate hazardous processes knows that a significant amount of the control strategy design is dedicated to how decisions are made when sensors fail.  Sensors used for autonomy are often redundant or overlapping.  But unlike computing a simple control valve command, autonomy must act on degraded - but not failed - sensors and decide a sophisticated planned action.

Vehicle Autonomy Replaces Drivers and Pilots

The military has been an early adopter for autonomy with aircraft and it appears that the days of the fighter pilot are numbered as new drone aircraft without the constraints of protecting humans replace pilots.  We already see a decline in commercial aircraft pilots and autonomy will accelerate that decline. In 2004, pilot error was identified as the primary reason for 78.6 percent of disastrous general aviation (GA) accidents  and as the major cause of 75.5 percent of GA accidents in the US.  A National Highway and Transportation Agency (NHTSA) study looked at the major accident causes and found that 2 percent of accidents were caused by the environment, 2 percent by the vehicles, and 2 percent came from "unknown" causes.  A full 94 percent, meanwhile, were caused by human error.

Implications of Vehicle Autonomy

Achieving vehicle autonomy will have profound changes for drivers/pilots, supply chain package delivery logistics, and for transporting people.  The cost of drivers and pilots are significant and transportation suppliers that can deliver services without those costs have a huge competitive advantage.  Autonomous vehicle service and support will surely require increased technical training and diagnostic tools for technicians. Early success in autonomy will have advantages, but there will be an ongoing battle for market share requiring safer, reliable, more efficient, more convenient, and cheaper autonomous vehicle solutions.

The technologies and intellectual property required for autonomy will create markets for new hardware, software, and services.  Casual conversations with your cab or Uber/Lift driver might be replaced with a virtual assistant as the vehicle identifies you at the start of your trip.  Perhaps passengers will tend to focus on reading, texting, working, and other applications as conventional seats are replaced with reclining seats with new infotainment and work-related devices used during transport.

Should low-cost autonomy become reliable and easy to use, there will be considerable economic incentive to move away from traditional individual car and truck ownership.  Autonomy would allow each transportation activity to adjust the vehicle to the task: a small, efficient vehicle for moving one person or perhaps a truck for a stop at the lumber yard.  The improving economics will eventually convince many to consider replacing at least one of their cars with autonomy (die-hard car buffs excepted).  This would be of great value to those that are not able to drive or choose not to drive themselves.

ARC Advisory Group clients can view the complete report at ARC Client Portal   

If you would like to buy this report or obtain information about how to become a client, please Contact Us     

Keywords: Certifiable Autonomy, Self-driving Vehicles, Automation, Navigation, GPS, LIDAR, RADAR, Situational Awareness, Artificial Intelligence, Machine Learning, Air Taxi, ARC Advisory Group.