A Beginners Guide to Autonomous Vehicles
Interest in autonomous driving technology is growing, though it’s often accompanied by widespread confusion. Let me clarify this topic by simplifying the various approaches. For a more comprehensive discussion, please refer to my book, Autonomous Vehicles: Opportunities, Strategies, and Disruptions.
Autonomous driving can be achieved through four distinct platforms, each employing unique sensors and software solutions tailored to different applications. The first necessary clarification is that there are four different ways to accomplish autonomous driving, which I call autonomous driving platforms. These are very different approaches to solving the autonomous driving problem, using various sensors and software solutions.
These platforms are not merely alternative solutions to the same challenge but are specifically designed to address different aspects of autonomous driving. Over the years, as the technology evolved, these distinctions became clearer, emerging from industry efforts to capitalize on particular market opportunities, which I categorize as follows:
Semi-autonomous: It primarily uses cameras to support advanced Driver Assistance Systems (ADAS) for highway driving in personal vehicles. This is classified as SAE Level 2.
Sufficiently Autonomous: Utilizes Lidar and high-definition maps to cater to autonomous ride-hailing, long-haul trucking, and delivery services. This platform corresponds to SAE Level 4.
Substantially Autonomous: More advanced than semi-autonomous, this may also incorporate Lidar and HD maps, generally fitting into SAE Level 3.
Fully Autonomous: Designed to operate independently in all conditions and environments.
Each platform aligns with specific autonomous driving applications and markets, reflecting autonomous driving technology’s diverse capabilities and goals. Later, I will explore how these platforms relate to different SAE definitions of autonomous driving.
Semi-Autonomous Driving
Semi-autonomous driving allows drivers to delegate specific tasks to their vehicles while still needing to remain aware by either touching the steering wheel periodically or keeping their eyes on the road. This technology has advanced from ADAS systems and includes automatic lane-centering, adaptive cruise control, and automatic braking.
Here’s how it works: When drivers enter a highway or a marked roadway they will drive on for some time, they activate the semi-autonomous feature (which varies by brand). They then set their desired speed relative to the speed limit, and the vehicle enters semi-autonomous driving mode. They must stay attentive and periodically touch the steering wheel to show they’re paying attention; otherwise, they receive a warning alert.
This technology relies on multiple cameras and radar sensors that feed information to an onboard computer to control the vehicle. While primarily for highway driving, it can also function on various road systems. However, semi-autonomous vehicles are restricted to forward driving and cannot make turns, placing them in the SAE Level 2 classification.
Currently, semi-autonomous driving is available in millions of vehicles and is regularly used to cover millions of miles autonomously on highways. I’ve driven over 10,000 miles using this technology in Tesla and Mercedes vehicles over the past four years. Below are examples of vehicles with semi-autonomous capabilities:
American Manufacturers:
Chevrolet: Malibu, Camaro (higher trims), Silverado 1500, Tahoe, Suburban, Colorado, Blazer, etc.
Cadillac: Various models
Ford: Mustang (higher trims), F-150, Explorer, Edge, Escape, Bronco Sport, etc.
Chrysler (Stellantis): Chrysler Pacifica, Jeep Grand Cherokee, Wrangler (higher trims), Dodge Durango, RAM 1500, etc.
Tesla: All Tesla models (Model S, Model 3, Model X, Model Y) come with Tesla Autopilot (ADAS features with limitations).
Asian Manufacturers:
Toyota: Camry, Corolla (higher trims), RAV4, Highlander, Sienna, Tacoma, Tundra, etc.
Honda: Accord, Civic (higher trims), CR-V, Pilot, Passport, Odyssey, Insight, etc.
Nissan: Altima, Sentra (higher trims), Rogue, Pathfinder, Armada, Frontier, Titan, etc.
Subaru: Legacy, Outback, Forester, Crosstrek, WRX (higher trims), Ascent, etc.
Hyundai: Sonata, Elantra (higher trims), Tucson, Santa Fe, Palisade, Kona Electric, etc.
Kia: K5, Forte (higher trims), Sportage, Sorento, Telluride, Niro EV, etc.
Genesis: G70, G80, GV70, GV80
European Manufacturers:
Audi: A3 (higher trims), A4, A5, Q3, Q5, e-tron, etc.
BMW: All BMW models typically offer a suite of driver assistance features (standard or optional).
Mercedes-Benz: All Mercedes-Benz models typically offer a suite of driver assistance features (standard or optional).
Volvo: All Volvo models are known for their advanced safety features, including ADAS functionalities.
Semi-autonomous driving vehicles are primarily marketed to upscale private buyers as an optional feature. However, they are increasingly becoming standard on high-end cars due to their potential to retain resale value as this technology gains popularity. While millions of vehicles likely have these capabilities, there are no reliable data on how many drivers use them regularly. However, it’s likely a small percentage as it takes time to get accustomed to using these features.
In summary, semi-autonomous driving is already a reality, powering millions of highway miles daily in the US. These vehicles cannot provide fully autonomous services as they require the driver’s presence. Full autonomous technology would enable vehicles to operate without any driver intervention.
Sufficiently-Autonomous Driving
The primary benefits of autonomous driving stem from what I term “sufficiently-autonomous driving technology” and the new market opportunities it generates.
Sufficiently-autonomous driving allows a vehicle to operate without a driver present but solely within predetermined routes and locations. This capability is restricted to specific geofenced areas defined by the company managing a fleet of vehicles for that particular area, although these areas are continuously expanding.
The technology behind sufficiently-autonomous driving incorporates lidar alongside cameras and integrates high-definition maps. This combination precisely locates the autonomous vehicle on the HD map, offering crucial information like traffic signs, restrictions, and turning radius. Initially, this approach was considered the safest method for implementing autonomous driving.
Sufficiently-Autonomous Driving Technologies
Sufficenty-autonomous driving technology platforms are consistent across multiple companies, as can be seen in this summary:
Notice that each technology platform uses high-definition maps with lidar to position the vehicle on the map. This is the essential difference from other autonomous driving platforms.
Sufficiently-autonomous driving is not suited for privately owned vehicles because of the limited driving areas. This technology platform is focused on autonomous ridehailing services (robotaxis), autonomous delivery, and autonomous long-haul trucking. These are the three most significant market opportunities for autonomous vehicles. Limiting autonomous vehicles to a geofenced area is acceptable in these markets.
Autonomous Ridehailing Services (ARS)
I expect ARS to be the most significant opportunity for autonomous vehicles. The cost advantages of ARS, as I describe them in my book, are far superior to ridehailing. I estimate the cost/mile to be down to $1.50, which is half the estimated price of ridesharing. ARK Invest estimates a market for ARS of just under $1 trillion when the price per mile is a little over $1. Initially, ARS will compete against ridesharing in selected metropolitan markets, then expand to more markets.
ARS is already proven to be a viable commercial service. Waymo, for example, has tens of thousands of riders and has more than 20 million miles of autonomous driving experience. Clearly, the technology works, and people are paying to use it.
This map highlights the currently approved area for Waymo’s autonomous ridehailing services (ARS). Between September and February, Waymo AVs carried over 530,000 passengers around the city, most frequently bringing passengers to job centers like the financial district, event hubs like the Chase Center, and other areas — especially those with difficult parking situations.
Autonomous Long-Haul Trucking
Autonomous long-haul trucking drives back and forth on highways on their geofenced routes from and to special depots located just off the highway.
The autonomous long-haul trucking market is close behind the ARS market. It’s now clear that the initial autonomous long-haul market will be launched in the southwest, primarily Texas, and will use a transfer hub model.
As my book describes, autonomous long-haul trucking is a huge market opportunity. It provides some significant strategic and cost advantages. It addresses the problem of driver shortages, significantly improves delivery times since trucks won’t need to sit idle during mandatory driver rest periods, and improves safety. There are also significant cost advantages that will drive initial use.
Several companies are already doing driverless shipments with a safety driver on board. They plan to remove the safety drivers later this year or early next year to start ramping up autonomous deliveries. Aurora currently makes about 100 weekly deliveries for FedEx, Uber Freight, and others. Aurora says it plans to have about 20 fully autonomous trucks working the 240-mile stretch between Dallas and Houston by the end of this year. Eventually, it plans to operate thousands of trucks all across America.
Autonomous Delivery
Autonomous food, grocery, and package delivery to homes and similar locations has tremendous potential. It is still being determined whether autonomous delivery will be provided by third-party services such as Uber Eats or DoorDash or by fleets of autonomous delivery vehicles operated by major chains such as Domino’s Pizza or Walmart.
Substantially-Autonomous Driving
Certain semi-autonomous vehicles are progressing towards greater independence, aiming to handle tasks like stopping at traffic lights and stop signs and executing left and right turns on their own. These vehicles could handle over 90% of driving autonomously with substantial autonomy. However, they will only be able to operate autonomously in some situations, and drivers may still be needed, albeit not immediately, for control. This falls under Level 3 in the SAE classification system. Substantial autonomy isn’t currently available, but some technologies are advancing toward achieving it.
Substantially-autonomous vehicles might also navigate autonomously without a driver in specific scenarios, relying on proven, map-based systems. For instance, a person might use their substantially-autonomous vehicle for a reliable commute to and from work. Alternatively, following a known route, such a vehicle could transport teenagers from school or sports practice without a driver. While theoretically possible, this concept requires validation. If proven, it could be referred to as SAE Level 3+.
Examples of semi-autonomous vehicles showing potential for substantial autonomous capabilities include the Mercedes-Benz Drive Pilot, GM Super Cruise, and Tesla FSD. However, FSD is uncertain due to its lack of reliance on detailed maps.
MercedesBenz Drive Pilot
The Mercedes-Benz Drive Pilot is the world’s first commercially available SAE Level 3 conditionally automated driving system. It can take over driving tasks under specific conditions, allowing drivers to look away from the road and remove their hands from the wheel for extended periods.
This system relies on several key technologies:
Sensor Suite: Drive Pilot integrates cameras, radar, and lidar to create a 360-degree view of the car’s surroundings.
High-Definition Maps: Unlike some other systems, Drive Pilot uses high-definition (HD) maps to provide detailed information about the road layout, including lane markings, traffic signs, and potential hazards. These maps are crucial for autonomous navigation within the system’s operational limits.
Operational Safety: The Drive Pilot software is designed to be very cautious. It can only be activated on pre-approved freeway sections during daylight and good weather conditions, without any construction zones. Additionally, the system has redundant mechanisms for critical functions like steering and braking to ensure safety.
While offering a glimpse into the future of highly autonomous driving, Drive Pilot currently operates at 40 mph or lower speeds in heavy traffic situations. It’s essential to note that Drive Pilot is still categorized as a Level 3 system, meaning the driver is legally responsible for the vehicle’s operation and must be ready to take control when prompted by the system.
As of April 2024, Drive Pilot is available through subscription for the Mercedes-Benz S-Class and EQS sedans in California and Nevada. Drive Pilot’s capabilities and availability may expand as testing continues and regulatory approvals are obtained.
GM/Super Cruise
Despite ongoing technological advancements, GM Super Cruise remains a Level 2 partially automated driving system. Key aspects of its technology include:
Technology Stack: Super Cruise utilizes LiDAR, radar, and cameras to perceive its surroundings. Notably, LiDAR contributes detailed 3D maps for precise lane positioning.
High-Definition Maps: Like the Mercedes-Benz Drive Pilot, Super Cruise relies on pre-loaded high-definition (HD) maps. These maps contain vital information about compatible highways, lane configurations, and potential hazards, enabling the system to operate effectively.
Driver Monitoring: Unlike some Level 2 systems, Super Cruise employs an interior camera to monitor the driver’s eye movements. This ensures the driver remains attentive and ready to assume control as needed.
Despite offering advanced features such as automatic lane centering and lane changing on suitable highways, Super Cruise is not classified as a Level 3 system like the Mercedes-Benz Drive Pilot. Consequently, the driver retains responsibility for monitoring the road and being prepared to intervene at any time.
Operation is restricted to pre-mapped highways deemed suitable for hands-free operation by GM. The driver must remain attentive and engaged even on compatible highways. Inattentiveness prompts the system to alert the driver to take control.
While not yet achieving Level 3 status, GM continually enhances Super Cruise technology. Expanding to additional roadways and advancements toward greater automation may be on the horizon, but it remains an advanced Level 2 driver-assistance time being.
Tesla Full Self-Drive (FSD)
Tesla’s Full Self-Driving Capability (FSD) is a suite of driver-assistance features that push the boundaries of Level 2 automation. However, it’s crucial to clarify that despite its name, FSD doesn’t achieve true self-driving capabilities (SAE Level 4 or 5).
Here’s how Tesla’s technology differs:
Sensor Suite: Tesla perceives its surroundings exclusively using a vision-based system with a network of eight cameras. This differs from competitors who integrate lidar and radar sensors for a more comprehensive view.
No High-Definition Maps: Unlike some Level 3 systems, Tesla FSD doesn’t currently use high-definition (HD) maps. Instead, it relies on its camera network and machine learning algorithms for navigation. This approach can limit its effectiveness in handling complex scenarios or poorly marked lanes.
Tesla FSD remains categorized as ally automated driving system despite its progress. Consequently, the driver retains ultimate responsibility for the vehicle’s operation and must be ready to take control as needed. Due to its reliance on cameras and the absence of HD maps, FSD can face challenges in certain conditions, such as adverse weather, unclear lane markings, or unexpected obstacles. Drivers must stay vigilant and prepared to intervene.
Tesla is actively enhancing FSD’s capabilities, which may involve incorporating additional sensors or adopting HD maps in the future. However, as of now, FSD remains a Level 2 system requiring continuous driver supervision.
Summary
Differentiating between sufficiently-autonomous and substantially-autonomous driving is crucial. Sufficiently-autonomous vehicles can operate without a driver but are confined to specific areas. On the other hand, substantially-autonomous vehicles necessitate a driver onboard. Substantially autonomous driving will likely feature in personal vehicles but not in the previously mentioned markets.
Even if substantially-autonomous vehicles can navigate selected routes without a driver, they still don’t achieve full autonomy.
Fully-Autonomous Driving
Fully-autonomous driving corresponds to SAE Level 5, allowing vehicles to operate anywhere and anytime without a driver present. However, viable fully-autonomous driving technology doesn’t currently exist and is unlikely to be achievable for a considerable period.
While Tesla’s Full Self-Driving (FSD) aims for this level of autonomy, it remains doubtful. The ability to autonomously navigate a vehicle in diverse environments such as dirt roads, narrow alleys, complex parking lots, detours, boat docks, fields, or traffic circles is currently beyond the capabilities of Tesla or any existing technology. Achieving this level of autonomy would necessitate a comprehensive smart infrastructure and high-definition maps covering every conceivable location, a development projected to take over a decade and requiring technology distinct from FSD.
Conclusion
In conclusion, the framework for autonomous driving is simple to understand. It consists of four different technology platforms for four different applications and markets. Two of them are currently operational and available, with one, sufficiently autonomous driving, the primary technology for creating a substantial new market for autonomous driving.
