On December 10, 2019, Phillip Goter and Joseph Herriges hosted the webinar “Autonomous Vehicles: Technical Advancements and Legal Considerations.” If you were not able to attend the webinar, you can find a partial summary of its contents in the Q&A below.
How Do Autonomous Vehicles (AVs) Work?
AVs work by detecting, recognizing, anticipating, and responding to the movements of transportation system elements without direct input from a human driver. Transportation system elements in this context include other vehicles, pedestrians, and cyclists, as well as the vehicle’s environment, such as roadway infrastructure, buildings, signs, pavement markings, and weather conditions. The safe operation of an AV requires connectivity between the vehicle and other elements of the transportation system. Engineers have identified six key types of connectivity:
- V2I: Vehicle to infrastructure (roads, traffic lights, mile markers, etc.)
- V2V: Vehicle to vehicle (other AVs)
- V2C: Vehicle to cloud (data to power navigation systems, entertainment systems, etc.)
- V2P: Vehicle to pedestrian (passive communication from pedestrian to the AV)
- V2X: Vehicle to everything (bicycles, buildings, trees, etc.)
AVs are enabled by artificial intelligence systems and connectivity. One example of an artificial intelligence system utilized by AVs is a neural network. A neural network is a software-based artificial intelligence machine that is modeled on the way the human brain analyzes information. AV neural networks use algorithms to detect patterns and classify and cluster data from the sensors in the vehicle. However, before they are fully operational, neural networks must be trained using a set of data that is labeled and classified by humans. In other words, AI systems must be “fed” data before they will work properly.
For example, consider an algorithm that the AV would use to identify stop and yield signs. An engineer would first create a data set of millions of images of different kinds of signs, making sure to capture the images during the day and night and from different angles to get a representative sample. The engineer would then go through all of those images and label them either as stop signs, yield signs, or something else. Once the data was classified, the engineer would get the neural network started by telling it what some of the images are (stop sign, yield sign, or other). The neural network would then be allowed to run through the data set on its own to classify each image. When it is done, the engineer would tell it whether its classification was correct. This process would then be run again repeatedly until the neural network achieves 100% accuracy.
What Is the State of the AV Market?
As with most emerging technologies, the AV market is in a state of expansion and rapid innovation. It is helpful to think of AV technology evolution on a spectrum. According to the Society of Automotive Engineers (SAE), there are six levels of driving automation:
- No Automation: Zero autonomy; the driver performs all driving tasks.
- Driver Assistance: Vehicle is controlled by the driver, but some driving assist features may be included in the vehicle design.
- Partial Automation: Vehicle has combined automated functions like acceleration and steering, but the driver must remain engaged with the driving task and monitor the environment at all times.
- Conditional Automation: Driver is a necessity, but is not required to monitor the environment. The driver must be ready to take control of the vehicle at all times with notice.
- High Automation: The vehicle is capable of performing all driving functions under certain conditions. The driver may have the option to control the vehicle.
- Full Automation: The vehicle is capable of performing all driving functions under all conditions. The driver may have the option to control the vehicle.
The vast majority of the United States vehicle market is SAE Level 1, with widespread adoption of vehicles that have safety and limited automated driving tasks, but that still require a human driver to exercise primary control of the vehicle. Achieving greater automation levels is driving growth in the industry, the value of which is expected to expand from $54 billion in 2019 to $577 billion in 2026. Patent filings for AV-related technologies at the United States Patent and Trademark Office (USPTO) also reflect this advancement, with filings increasing 20-30% per year since 2016. Looking ahead, industry analysts predict that vehicles with some self-driving capabilities likely will be available in the early 2020s, and that fully autonomous vehicles may arrive in the late 2020s.
What Are the Challenges of Developing AV Technology?
The primary challenges facing the development and implementation of AV technology are the variability of both drivers and the environment. Interactions between AVs and traditional vehicles (or “non-connected vehicles”) pose a particular challenge, as non-connected vehicles lack the V2V communication systems that help AVs avoid collisions. To compensate, AVs must undertake the difficult task of attempting to predict driver behavior.
Transportation infrastructure also poses a challenge due to its lack of uniformity and the frequent environmental and structural changes it undergoes. For example, AVs must be taught to recognize that some roads use painted lines to mark their center points, while others do not. They must also learn to adapt to changes, such as recognizing a new painted roadway marking even when the old marking is still partially visible. And they must recognize transportation infrastructure in inclement weather, such as rain and snow.
What IP Issues Could Companies Face?
Patent eligibility issues are front and center as IP challenges for AV companies. Under § 101 of the Patent Act and the Supreme Court’s Alice and Mayo decisions, the USPTO may reject applications for software-implemented technologies for tasks that previously were performed by humans, deeming these inventions unpatentable abstract ideas under the current state of § 101 law. The challenge practitioners face is in drafting claims that include “something more” than the abstract idea or that recite an “inventive step” – a sometimes-nebulous standard that may raise a difficult hurdle for applicants to clear. Many AV and AI technologies are software-based, making them vulnerable to such rejections. Technologies that replicate what a human mind would do – such as neural networks – are particularly difficult to patent, and require carefully drafted claims to pass the Alice/Mayo test of patent eligibility.
AI technology also raises questions of patent infringement liability, as it is not yet clear how the legal system would handle infringement by AI systems. In this context, the actual “infringer” may be the AI system itself, the AI system’s original coder, its owner, or any other entity that trained or operated it.
What Are the Trade Secret Issues in the AV Space?
Trade secrets can serve as an effective alternative to patent protection, as they do not require lengthy and expensive prosecution, nor do they expose their owners to patent challenges. A trade secret is any confidential business information that derives economic value by virtue of being unknown by others, and that the trade secret owner has identified internally as such and taken reasonable efforts to maintain as a trade secret. Such information can include source code, research and development, designs, prototypes, processes, and marketing plans, among others. To establish the existence of a trade secret, the proponent must show that the information:
- is valuable;
- is not known by others who might profit by its use; and
- has been handled by means reasonably designed to maintain secrecy.
Trade secrets may prove particularly useful in the AV space to protect negative know-how – i.e., knowledge of what doesn’t work rather than of what does work. Negative know-how is especially valuable in emerging industries such as AVs, as it can give the company that owns it an R&D advantage over competitors. However, maintaining IP as a trade secret rather than as a patent is not without risks. The ease with which data can be transferred increases the likelihood of employees stealing or leaking trade secrets, and it can be difficult for plaintiffs in trade secrets cases to demonstrate misappropriation.
What Are the Products Liability Considerations?
Collisions involving AVs pose unique questions of liability. Traditionally, liability for collisions between non-AVs was primarily based in the law of negligence, with products liability issues occasionally arising if the collision was caused by a defective product (e.g., faulty brakes, tire blowouts, etc.). Products liability law is likely to play a greater role in determining fault in collisions between AVs because these collisions may be more likely to be the result of software glitches or mechanical failure, rather than human error. The extent to which products liability will supplant negligence as the primary method of allocating fault will depend upon the degree of the vehicle’s automation, as negligence could still play a role where a human driver exercises some measure of control over the vehicle.
What’s Currently Happening with Federal Legislation?
As of the date of publication, there has been no specific federal legislation aimed at regulating the AV industry. However, 37 states have enacted AV-related regulations through legislation, executive order, or both. Many of these state-level regulations permit AV companies to use the state (or a specific location therein) as a testing ground.
For more information about autonomous vehicles and artificial intelligence, as well as the law affecting both, please see “Autonomous Vehicles: Technical Advancements and Legal Considerations” or contact Phillip Goter or Joseph Herriges.