Much has been written about the use of ‘automated’ versus ‘autonomous’ when it comes to robots and vehicles (car, trucks, etc). Despite that, I’ve not seen a short & clear analysis of these key words based on true definitions. This post attempts to be that analysis and clarify the distinct nature between these terms.
The Society of Automotive Engineers (SAE) has published J3016 : Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles. The 35 page PDF is free to download, but does require registration with SAE. In it, they decree autonomous, self-driving, and unmanned as a incorrect terms and recommend the term ‘driving automation’ or ‘automated driving’ instead.
I do not agree with their recommendation and find fault in their supporting text.
They do recommend that their preferred term be applied to the action of driving, instead of the physical vehicle itself. On this point I have no strong opinion, and I agree that it provides clarity when a system has available features that are not currently active or engaged.
Automated?
In Section 7.1 of J3016, the authors write:
Vernacular terms such as [Autonomous, Driving Modes(s), Self-Driving, Unmanned, Robotic] are sometimes used—inconsistently and confusingly—to characterize driving automation systems and/or vehicles equipped with them. Because automation is the use of electronic or mechanical devices to replace human labor, based on the Oxford English Dictionary, automation (modified by “driving” to provide context) is the appropriate term for systems that perform part or all of the DDT. The use of other terms can lead to confusion, misunderstanding, and diminished credibility.
Tracing this along, the Oxford English Dictionary (OED) actually defines automation as:
- The use or introduction of automatic equipment in a manufacturing or other process or facility.
and automatic as:
- (of a device or process) working by itself with little or no direct human control.
- Done or occurring spontaneously, without conscious thought or attention.
And there-in lies my issue with J3016 – both definitions of automatic are valid for automation technology. Equipment like dishwashers, printing presses, robot arms in factories, thermostats all:
- Automate a process
- by itself, with little or no direct human control
- AND are done spontaneously, without conscious thought or attention.
I strongly argue that autonomous mobile robots (including self-driving cars) do NOT fall into that definition. Yes, technology applied to these domains does reduce human labor. But critically, actions are not done spontaneously or without attention. In these systems there is a staggering amount of computation devoted to perceiving the surrounding world and then deciding what actions are safe & effective toward the desired goal.
Autonomous?
In Section 7.1.1 of J3016, the authors write of of the term ‘autonomous’:
This term has been used for a long time in the robotics and artificial intelligence research communities to signify systems that have the ability and authority to make decisions independently and self-sufficiently. Over time, this usage was casually broadened to not only encompass decision making, but to represent the entire system functionality, thereby becoming synonymous with automated. This usage obscures the question of whether a so-called “autonomous vehicle” depends on communication and/or cooperation with outside entities for important functionality (such as data acquisition and collection). Some driving automation systems may indeed be autonomous if they perform all of their functions independently and self-sufficiently, but if they depend on communication and/or cooperation with outside entities, they should be considered cooperative rather than autonomous. …
Additionally, in jurisprudence, autonomy refers to the capacity for self-governance. In this sense, also, “autonomous” is a misnomer as applied to automated driving technology, because even the most advanced ADSs are not “self-governing.” Rather, ADSs operate based on algorithms and otherwise obey the commands of users.
Note that I’ve added the emphasis in this excerpt, it does not appear in the original text.
The authors argue that autonomous is not the proper term because autonomy / autonomous-operation is defined as a system which operates entirely standalone, without communication or cooperation with other entities.
The OED does support their definition of autonomous:
- (of a country or region) having the freedom to govern itself or control its own affairs:
- Having the freedom to act independently.
- Denoting or performed by a device capable of operating without direct human control.
but not their conclusion.
SAE argues that ‘autonomous’ should not be applied to vehicles which require coordination or communication with outside entities. The grouping of “coordination required” and “communication required” is a red herring. I argue that outside data collection does not violate 1.1 of the OED definition, and that freedom to act independently still allows retrieval of weather, traffic, map information, GPS position, etc from an outside sources (instead of directly observing those conditions). Do I become non-autonomous when I check the weather forecast when I decided what to wear in the morning?
The SAE also states that since these systems follow the rules of their algorithms and obey direct control by users that they are not self-governing and therefore are not autonomous. If this interpretation were applied to another system – say the US Federal Government – I guess the SAE would conclude it not to be autonomous either as it:
- Coordinates and communicates with other countries.
- Follows legal doctrine in its operation.
- Obeys the outcomes of elections (user-control).
Of course, the US Federal Government (like all other governments) is autonomous and SAE’s bizarrely-strict interpretation of the term is plain wrong. In short:
- The freedom to act independently does not preclude one from interacting or coordinating with others.
- The nature and location of decision making has far more bearing on a system being autonomous (or not) than if said system fetches or observes all input data.
- An entity that follows rules can still be autonomous as rules set the boundaries or limits of possible actions – rules do not deterministically define the next action based on the current state.
- Being capable of operating without direct user control does not preclude listening to direct control when it is provided.
A system is autonomous when it has the capability of achieving the points of the OED definition. If said system chooses from time to time to ignore its’ freedom and play nice with other – it is still an autonomous system.
The OED definition also separates these sorts of advanced mobile robots & vehicles from “automation technology”. It is impossible to argue that a dishwasher governs itself or that a factory robot arm operates independently.
Conclusion
Autonomous is the correct term to apply to these kinds of advanced mobile robots and vehicles.
Note that a system described as automated or autonomous does not relate or correlate to its’ effectiveness or safety. The vehicles being developed and tested by Argo, Aurora, Uber, and many automakers are best described as autonomous – but I would not currently label them as effective, reliable, or safe. These systems are in development and there are many technical hurdles to overcome before fielding.
I agree with the SAE that the terms ‘automated’ and ‘autonomous’ are often used incorrectly and that leads to confusion in the marketplace and in the legal world. We should all strive to be more accurate and precise with our language uses. It’s especially difficult to have clear and non-overlapping definitions of capabilities that exists on a continuous spectrum. But, labelling an autonomous system as automated (or vice versa) results in incorrect assumptions about its’ design and will increase confusion in the marketplace.
This article was published on 2018-09-14