As the hype around driverless vehicles continues to sow the seeds of overexpectation, driver future job concerns, and less than realistic timetables, it occurs to me that there is a need to consider a basic point: How do we prove a vehicle is truly driverless?
In earlier blogs, I’ve highlighted the concern that demonstration of a technology does not mean the sale of said technology is in the near offing. A lot goes into proving a technology is capable before it can be commercialized. Watching a video of a technology in action may look very impressive, until you learn that the road was closed for the test and the vehicle was surrounded by test drivers ready at a moment’s notice to swerve should something go awry in the demo.
Don’t get me wrong – demos are an important way of understanding the capabilities of a technology. But context is critical and knowing the details around a demo are important to truly understanding what the technology can really do … and what it cannot.
Plus, you don’t get to ask questions of a video, the same way you can at a demo!
So, how will we know when a vehicle is truly driverless?
To answer this question, I propose what I call simply “The Test.” When a driverless vehicle can do what I propose, then I will be convinced that driverless technology is, or is truly soon to be, here.
Since we’re in the trucking industry, we’ll focus “The Test” on the trucks. However, the same test could be used to prove driverless car technology as well. I’m starting with trucks because I believe trucking has a lot more to gain (and drivers a lot more to lose) when driverless technology is applied. Think of what other autonomous technologies have done to industries and the jobs in them, and you’ll know what I mean. (Remember tollbooth operators ... with license plate scanning and automated toll collection, projections are, these jobs will soon be gone.)
Today, the most advanced driverless technology will not pass “The Test.” Oh sure, there are cars driving around communities relatively freely – like the Waymo cars in Mountain View, California, and the Uber cars in Pittsburgh. And we’ve even heard recently of driverless taxi service coming soon to a community near you. But hop in one and ask it to take you to Grandma’s in Denver and it will get you to the city limits and then ask you to take over! (Unless of course, you’re in Denver!)
These are what we call “Level 4” technologies – purely limited autonomy. In other words, as long as the vehicle is traveling with its defined operational design domain (ODD), it can drive autonomously. However, add something outside its ODD – like going outside its currently mapped boundary (Mountain View or Pittsburgh), or inclement weather instead of sunshine – and the system looks to the driver to take over. To me, limited autonomy is not driverless – it’s still automated. And, it still needs a responsible driver behind the wheel to ensure the vehicle is able to arrive at its destination whenever and wherever it needs to.
Level 5 is full autonomy, or truly driverless. The technology will get you where you want to go, wherever that may be and whenever you want to go – no matter what!
So let’s talk about “The Test” for truly driverless (Level 5) trucks. Here’s what I’m proposing:
- In order to prove that a vehicle technology is truly driverless, the following must happen:
- A tractor-trailer or truck, which is equipped with no driver controls (steering wheel, pedals, shifter), is fully loaded (GVCW = 80,000 lbs) and can travel from Bangor, Maine to Los Angeles, California, deliver its load, be reloaded, and make the return trip, in equal to – or less – time than that of a driver team carrying a similar load.
- Some additional points:
- The test must be conducted in late January or February to ensure the technology is robust and viable in inclement weather.
- Note: The vehicle may pull off the road to wait out truly treacherous conditions as defined by actual weather warnings, such as tornado or blizzard warnings, or can determine independently that conditions are beyond its span of control and pull off the road to wait. However, the vehicle itself needs to determine that conditions warrant a safe haven.
- There can be no human input – either within or outside the vehicle – to support the autonomous operation, or to make any decisions impacting the vehicle’s route or other aspects that would render the vehicle under partial human control.
- The route must be the most reasonable path available and both vehicles (the driver team or the autonomous vehicle) must travel the same route. Sorry, no shortcuts south to warmer climates and skedaddling in the sunshine. However, a different reasonable route must be used for the return.
- The vehicles must follow all traffic laws and regulations for each of the states traveled. This includes pulling into weigh stations (or using a pre-pass system).
- The automated vehicle cannot be involved in any crashes – whether its fault or the fault of others. It has to be “smart enough” to anticipate and react to potential problems created by human drivers who share the road.
- Both vehicles (again, the driver team and autonomous truck) must travel normally in all types of traffic conditions. Neither can be surrounded by escort vehicles.
- Vehicles must be driven 24 hours a day, unless weather – as noted earlier – or any other malfunction requires it to be sidelined. (Human-driven vehicles must follow current Federal Motor Carrier Safety Regulations regarding HOS, though an exemption can be made for the test, or three drivers can be used in the vehicle.)
- Humans can load the automated vehicle, as well as refuel, change tires, or perform maintenance as required. However, the autonomous vehicle must be able to alert humans when these items are needed.
- Connected vehicle technology can be used, but there can be no connectivity that enables the vehicle to be driven remotely – such as by a lead or trailing vehicle, or from a remote location. The one exception is if there is a need to move the vehicle off the road due a malfunction that the system is not able to correct. Therefore, the vehicle can be monitored, but normal driving situations cannot be controlled.
- If the vehicle needs a repair enroute, one that it cannot diagnose and self-heal, then a human team can do the repair and the vehicle can continue on the test. However, as required by current guidelines, the vehicle still must be able to move itself to a “safe state” and not inhibit traffic or cause an accident.
In a nutshell, the vehicle has to do the things a human driver would have to do to run a route, deliver the goods, load up, and head home. When an autonomous technology can pass this “Test,” then, at least in my mind, it’s truly driverless.
When this happens, autonomous driving will have arrived. (Personally, I don’t think there’s a vehicle that will pass this test in the next 40-50 years. But, if I were a successful prognosticator, I would be enjoying the warmth of a sandy beach somewhere – having made millions in the stock market or the lottery – rather than reporting to the office each day!)
But, until this happens, let’s stop calling these technologies autonomous and call them what they really are – automated.
Technical and industry insight from OUR experts.