Five years ago, it marked a peak in the exaggerated cycle of “self driving” cars. At that time, almost all major motor vehicle manufacturers and hi-tech companies predicted that the automatic driving system (ADS) would be widely deployed in 2020, which is believed to lead to the rapid elimination of traditional human driving.
In hindsight, the mainstream view in this period was obviously wrong. Up to last year, there were few advanced prototype vehicles running on public roads, and there was no need for on-board safety driver intervention when the automation system needed manual help. The term “automatic driving” has lost its original meaning because the driving assistance functions on vehicles marked as “fully automatic driving” cannot be manipulated without continuous manual supervision. The relevance of “automobile” to today’s automation is far less than that of trucks, buses and passenger and cargo vehicles.
By 2018, the CEOs of major companies (waymo, general motors, Ford and Aurora) who invested most in advertising began to make public statements, pointing out that the introduction of automatic driving will be gradual, which eased their previous optimism, starting with the operation under the constraints of strict location restrictions. At their current pace, it will take decades to expand to any place close to national deployment. The organizational learning curve and cost are much longer and higher than expected. After investing at least a decade and billions of dollars in advertising development, these companies learned that the technical requirements to support the widespread use of the technology were far more complex than they originally thought. At the same time, companies such as Tesla and less mature start-ups that continue to accelerate the deployment speed and expand the deployment scale are still learning hard and have not realized how far away they are from the goal.
Knowledgeable advertising developers have reached a consensus on several key aspects of the technology and its recent deployment:
In the next few years, automation will be feasible only under narrow conditions, including good weather, lighting, traffic and electronic geofencing locations that have been mapped with high accuracy and, in many cases, equipped with appropriate physical and digital infrastructure support functions.
Since ads needs to avoid all traffic hazards encountered without manual driving intervention, these systems must include a higher level of safety assurance than driving assistance systems relying on manual supervision.
Advertising needs to rely on multiple independent sources of information about the driving environment and its hazards. These data come from cameras, lidar (light detection and ranging system), radar and precise positioning, as well as highly detailed digital maps.
Although many advertising developers claim that they can drive without relying on wireless communication from other vehicles, and without relying on warnings from vulnerable road users and the actual roadside infrastructure itself, recent studies have shown that the widespread use of advertising without such cooperative communication may have an adverse impact on traffic flow, energy use and environmental emissions, Because it is impossible to predict future road conditions.
The technology will initially be used for specialized purposes, such as local parcel delivery, long-distance truck transportation on highways, urban transportation services on fixed routes, and automatic passenger ride calls in cities and suburbs at more limited locations.
Even if ads can drive the vehicle without on-board manual driving as a backup, they still need the remote support of skilled drivers to manage “emergencies” that cannot be handled by automation.
These conclusions come from some basic experience learned by technology developers. Most importantly, the inherent complexity of driving tasks makes it difficult for automated systems to accurately perceive the driving environment, predict the behavior of other road users, and identify and deal with traffic hazards.
Just as children learn to control their movements, advertising needs to learn to crawl first, then walk, and then run. This is why these functions must first be implemented and improved in a simple environment before dealing with complex interactions with unpredictable road users, including pedestrians and cyclists, or operations under adverse weather conditions, such as heavy rain, snow, fog and icy roads.
Ads needs to perceive the surrounding environment and determine its location using technologies based on a variety of basic physical principles to provide safe operation. They need to deal with the adverse conditions caused by electromagnetic interference from electric storms or nearby electrical equipment, low solar angle that may lead to camera blindness, precipitation or smoke of light required by diffusion imaging sensors, and cyber attacks against any vehicle sensor technology. Information is also needed on the relative speeds of nearby infrastructure and other moving objects in the vicinity. Data from all these sensors must be fused to accurately represent the area around the vehicle and isolate any fault inputs.
In the foreseeable future, sensor requirements and data processing and storage related to this environmental sensing level will make ads technology expensive. If this technology is to replace all the skills of drivers without affecting safety, these costs are inevitable.
These costs also drive the business case for the first deployment of ads by commercial fleets that can be used all day to generate revenue, rather than running for only one to two hours a day, as most privately owned vehicles do. While automation technology is still improving, commercial fleets provide further advantages. It will be easier to properly maintain sensitive components on the fleet, and fleet operations will help provide remote support to vehicles when needed.
Remote manual support for autonomous driving is an important topic, but it has not received due public attention. In fact, every advertising developer without on-board manual driving hopes to rely on the manpower of the team management center to help the system make tactical driving decisions when needed. The human assistant can compensate for hardware and software failures on the vehicle, but mainly provides advice in challenging traffic situations, such as navigating around obstacles temporarily blocking the vehicle’s scheduled route, determining when traffic rules may need to be violated to deal with specific strange situations, identifying police gestures to direct traffic, and identifying objects in the vehicle path. Some developers are also considering the remote driving function, in which the personnel of the fleet management center will drive the vehicle in some journeys that cannot be handled by ads. With the continuous improvement of technology, these remote support functions will be particularly important in the early stage of ads deployment. They will also provide employment opportunities for many drivers whose jobs will gradually be replaced by advertising. The working conditions of remote driving may be more attractive than traditional long-distance truck driving, allowing drivers to work near home, and they can avoid the physical discomfort and safety risks associated with traditional truck driving.
Although the most popular intended use of autonomous driving a few years ago was automatic call passenger service, interest has clearly shifted to the automatic transportation of goods – autonomous heavy trucks on long-distance interstate routes and small low-speed vehicles transporting packages in local cities and suburbs. This shift of focus started before the pandemic but accelerated during the COVID-19. In dense urban areas, taxi services have the most favorable demand profiles, short trips and shared rides. These places are the most challenging environments in ads technology. In contrast, the long-distance truck transportation business on the interstate faces simpler traffic and road infrastructure conditions, especially if the starting and ending points of automatic travel are directly connected to the limited access highway. Urban and suburban parcel delivery can use vehicles that are too small for people, and can drive at low speeds on sidewalks or bicycle lanes. Freight vehicles have additional advantages. They can be designed to drive with extra care to avoid conflict with other road users without worrying about passengers’ impatience or the need to protect their passengers from injury.
Not every advertising company has adopted this method. In terms of advertising deployment, Tesla is an obvious outsider compared with other companies in the industry. It pursues the traditional automobile industry model, sells privately owned personal vehicles to meet all purposes, and tries to improve its ads capability by building its current driving assistance system rather than designing a system with the necessary higher redundancy and fault management level. The company has recently focused on methods that rely entirely on machine vision, avoiding the use of supplementary sensors or accurate maps and positioning that provide key information around the vehicle.
Tesla’s approach reduces production costs, but seems to hinder safe and reliable performance in the face of inevitable changes in traffic and environmental conditions. The company uses the names of “autopilot” and “fully automatic driving” in its existing products, which makes the public seriously confused about the functions of Tesla products. What is even more puzzling is that the owner’s Manual of Tesla and the letter from the government safety regulatory agency described the technology as a driving aid system rather than an automatic driving system. In fact, Tesla’s use of the name “fully automatic driving” has created a serious public misunderstanding, so that Alphabet’s Waymo, originally known as Google’s autopilot project, announced that it would no longer use the word “autopilot”.
While dealing with various traffic hazards and internal hardware and software failures, ensuring the safety of ads operation remains the main technical challenge of deployment. Ads researchers and developers have studied a variety of security assurance methods, but they have not reached a consensus on the best scheme. The challenge of security assurance goes far beyond technical considerations, but the first is broader social decision-making, that is, how to determine the “security” of advertising in public services. Relevant security measures need to be agreed so that regulators and the public can adapt to public deployment. Advertising developers also need to learn how to display their highly technical safety assessment in a way that regulators and public interest groups can clearly and accurately understand, so as to win trust and recognition.
Although some observers may think that in the Current hype environment, the “prosperity” of automatic driving is over, the current situation actually marks a sign of progress. A more realistic view of the opportunities and challenges of autonomous driving will promote more targeted investment in resources and align public awareness with reality. We expect that in the current decade, automated long-distance truck transportation on low-density rural roads and automated local small package transportation in urban and suburban environments will be limited. Automatic car Hailing services in cities and suburbs can also be provided on a limited basis, but the location-specific challenges faced by their deployment are enough to make it impossible to reach a national scale soon.