The Law Commission in the United Kingdom recently completed its massive study on domestic legal reform for automated driving. As the UK government works to implement the study’s thoughtful recommendations, the Commission’s experts are now turning to the topic of remote driving. I’m happy to offer a few thoughts.
First, “remote driving” encompasses a range of scenarios.
The remote human might be:
- next or even inside the vehicle (as with some parking assistance features)
- in a nearby moving vehicle (as with some early platooning concepts)
- far beyond line-of-sight (and maybe even in another country—though Florida says no)
The remotely driven vehicle might be:
- in a private facility such as a dispatch yard or distribution center
- in a quasipublic facility such as a parking structure
- on a public road
The remote driving might involve:
- performing the entire dynamic driving task on a sustained basis
- supervising the performance of an automated driving system that is being tested
- performing fallback for an automated vehicle by bringing the vehicle to a minimal risk condition
- moving a stationary automated vehicle following a crash or other incident
Second, this “driving” involves performing all or part of what SAE J3016 calls the “dynamic driving task”—monitoring the driving environment and braking, accelerating, steering, and signaling as needed.
Third, remote driving is therefore more than mere “remote assistance.” Where exactly to draw the line between remote driving and automated driving has been contested, mostly by automated driving developers seeking to distinguish remote assistance (on which they generally rely) from sustained remote driving (of which they are generally skeptical). Under SAE J3016’s resulting definitions, the key question is whether a remote human actually “performs the dynamic driving task” by monitoring the driving environment and braking, accelerating, steering, and signaling as needed.
Imagine, for example, a vehicle that unexpectedly comes upon a construction zone with a travel lane marked by ambiguously placed traffic cones. A human who remotely steers the vehicle between these cones is facilitating remote driving, whereas a human who remotely sketches an approximate travel path that the vehicle then follows using its own sensors is merely providing remote assistance. (A separate question with no clear answer: If the vehicle waits in an active travel lane for this input, what level of automation does this represent?)
Fourth, automated and remote driving may overlap in practice—at least at the margins. On one side, a remotely driven vehicle may need some automation capability to either safely proceed or safely stop if its connection to the remote human is disrupted. On the other side, a first responder might direct the company managing an automated driving system to move an automated vehicle away from a crash site, which depending on the company’s approach and the vehicle’s condition might involve remote driving rather than mere remote assistance.
Fifth, at least in my view, the technical distinction between automated driving and remote driving is less relevant to the regulation of these forms of driving. I have long argued, and both the Uniform Law Commission in the United States and the Law Commission in the United Kingdom largely agree, that companies are the “drivers” and “operators” of their automated vehicles. The company behind an automated driving system drives via its machine agents (namely the combination of hardware and software that enables automated driving) and its human agents (namely the remote assistants who support and supplement that hardware and software). Indeed, while “automated driving system” in J3016’s narrow technical sense refers only to that hardware and software, in a broader sense this “system” might also properly encompass these human roles.
How exactly a company carries out this driving is a secondary question that can be reasonably answered in many ways. A company might primarily rely on its automated driving system (with remote human assistance), on its remote human agents (with automated safety features), or on some simultaneous or sequential combination of the two. Parsing whether, for example, the humans remotely facilitating the dynamic driving task are driving or merely assisting distracts from the more important question of whether the company itself is driving safely through its hardware and software, its employees and contractors, and its connections between them.
The drafting committee for the Uniform Law Commission’s Uniform Automated Operation of Motor Vehicles Act initially considered a draft that embraced this approach by declining to distinguish between automated driving and remote driving and therefore applying its provisions to both. Ultimately, however, the committee determined remote driving to be out of scope.
Sixth, remote driving within line of sight is a special case that could be appropriately conceptualized in two different ways. Consistent with my argument above, one might conclude that the company providing this feature uses the human on the ground to drive the vehicle. But given the particular proximity of that human, one might instead conclude that the human uses the company’s feature to drive the vehicle.
Seventh, regulating companies as drivers does depart somewhat from the 20th century approach to “driving.” On one hand, the person behind the wheel of a conventional delivery truck or taxicab is its legal driver—not the owner of the vehicle or the passenger in the backseat. On the other hand, if that driver is an employee acting within the scope of their employment, then their employer is civilly liable for harm caused by their negligent driving—though this principle has been undermined by companies’ putative use of “independent contractors” rather than employees.
Eighth, we should absolutely depart from the 20th century approach to driving! The safe systems approach, as modeled by Sweden and embraced by other countries, emphasizes that safety cannot start and end with the person behind the wheel—wherever that wheel may be.
Ninth, I am not aware of any extended analysis of how, in the United States, the federal motor vehicle safety standards (FMVSS) apply to remote driving. Half of this answer—whether these standards are consistent with the design of a remotely driven vehicle—evokes similar analyses for automated driving. The other half—whether these standards are applicable to and if so consistent with the design of any off-vehicle equipment used for remote driving—is more novel. But these issues do not seem to have arisen for the remote parking features that are already available on some production vehicles.
Finally, some US states have at least contemplated one or more of the forms of remote driving in statutes that are principally focused on automated driving. In fact, the United Kingdom’s Law Commission has done a fine job identifying many of these references. Given their modest context, however, these references do not represent the kind of policymaking that offers lessons for other jurisdictions interested in affirmatively, deliberately, and holistically regulating remote driving.