Was Thatcham’s Tesla Autopilot Test Fair, All Things Considered?

Even if you follow Tesla’s instructions with regards to their Autopilot technology to the letter, could your car still get into an accident during perfect weather and road conditions?

That’s something vehicle safety experts at Tatcham Research wanted to find out, which is why they devised an experiment that would put a Tesla Model S in a rather difficult, yet perfectly plausible emergency situation.

The Tesla performed as expected during an initial test, which involved coming to a complete halt after the car in front started braking. Autopilot was engaged and the Model S did exactly what it should have done with no input from the driver.

It’s the following test that proved problematic for the all-electric saloon, when journalists wanted to see how the Model S would react if the car in front suddenly changed lanes in order to avoid a stationary vehicle. While the Tesla did start braking, it did not change lanes so as to avoid the stationary obstacle.

Even though it takes more than just one test to talk us into this type of scenario being an Achilles Heel for Tesla’s Autopilot, we wouldn’t blame you if you thought it was something worth worrying about.

However, questions about the test’s validity have also been raised, where CarTests points to how the system in this particular Model S may not have recognized its surroundings as being a multi-lane road – possibly because the car only traveled a few hundred meters before the impact. Also, the sat-nav system may not have even registered that test track as a real roadway.

If either of these possibilities turns out to be a factor here, then it would be only logical for the Model S to not steer clear of the object ahead, seen as how that would basically mean running itself off the road.

As for scanning the road ahead and realizing there’s a stationary car there even before the vehicle ahead changes lanes, that’s a whole different story. What’s certain though is that semi-autonomous technology is far from what anybody should consider a “finished product”. So legally speaking, the driver of the Model S should have been the one reacting to this situation and applying the brakes – not the Autopilot system.

  • Matthew Boyd

    When they bump up to lvl 3 standards and at least attack LIDAR to their vehicles, then they can prevent these types of accidents. You can only go so far with a long range camera, radar and ultra sonic sensors.

    • D3X

      Exactly, but I think we should skip Level 3 entirely. All Automakers should skip and go right to Level 4, simply put; Putting Level 3 into consumer level cars will increase collisions due to improper use of the system. There’s already a lot of idiots on the road with Level 2 and simply not using the technology properly (sitting at the Passenger seat, texting, sleeping) and not keep their hands on the wheel at all times.

      A Volvo engineer back in 2016 mentioned this.

  • Status

    AutoPilot is, at best, a lane keeping and adaptive cruise system.

  • Ron

    I think the two glaring problems for Teslas ‘autopilot’ is the marketing behind the product and the ignorance of the consumer, to be frank. The very definition of the term is “a device for keeping an aircraft on a set course without the intervention of the pilot”. To me, thats not what Tesla has in its vehicles and its conveying the wrong message to the masses. Add that to the numerous youtube videos of Tesla owners driving down the road in the passenger seat, or watching a movie while driving as if their car is literally driving itself (when it clearly cant), what you end up with is a false sense of confidence and a complete misunderstanding of the cars abilities.

    • Ed Woodrick

      Yea, no. An autopilot in a plane is not a set and forget. You’ve still got to be looking for other traffic and listening to air traffic control. Also, by law, above 15,000 ft, you are always under positive control of ATC and no other traffic should be where you are.

      But you are right, it is a device for keeping an aircraft on course without intervention. That’s EXACTLY what the Tesla device does. The issue is that you CAN’T stay on a single course on roads. You’ve got to continually look out for things that change your course.

      • Stephen G

        After reading your post and Ron’s, I wonder why this feature is relevant. Adaptive cruise sounds beneficial, if it functions as expected. But Active Lane Control…if I’m paying attention I don’t need this.

  • jbmadness

    If the test arena wasn’t “safe” or recognized and the systems rely on know data on roadways then why would they allow the remote systems to be used in those conditions. It would be pretty easy to argue that this stuff isn’t ready for being on public roads yet.

    • Matthew Boyd

      That’s a good point. The car should track under all conditions versus set parameters.

  • LaszloZoltan

    certainly it happens worldwide all the time with varying results- comparison (better/worse) is meaningless because the object of this exercise is to avoid a collison

  • LaszloZoltan

    the electric motor on a tesla ought to be able to reverse in a split second which could provide some hefty braking power when required.
    with such energies involved I can imagine the physical dynamics wont be just a simple straight-line braking, but again, motion sensors might be used to provide data needed to modulate individual wheelspeed for better control

    overall, I am not concerned, humans are evolved problem solvers, we got this

    • Stephen G

      Braking is a problem too. It is too frequently used to avoid a collision (or imagined collision) and ends up causing one.

    • Ed Woodrick

      Going into reverse would spin the tires, but if you are saying 60-0 in less than a second, that’s also called a crash.

  • We’ll no, I don’t think it was. The ‘car’ was stationary when the Tesla hit so it had, at best, 30 to 40m to come to a complete stop. I don’t think any human would’ve been able to do it quicker, regardless of what car they were in.

    Also, this feels like a very unique situation as the car changing lane would’ve most likely slowed-down first because when is the outside lane that clear?

  • Bash

    OR…. the autopilot is SO smart that it knows that that was just a balloon and intentionally decided to ignore it.
    .
    PS: Mr. musk, where is my Cheque? 😀

  • LeStori

    I suspect many people, once they start braking, find it hard to change lanes as well. The normal reaction would be to brake harder, which is why ABS is used in modern cars. Part of the problem is that once you start braking, your eyes are fixed on the vehicle in front. To change lanes you need to deternine if there is somehting beside you or passing you in some way. This would require you to take your eyes off the vehicle you are about to hit . Not an easy thing to do for most people. Much easier to change lane suddenly if you are not braking.

  • Motoring Nation

    Tesla need to rebrand their ‘Auto Pilot’ system and stop with the marketing that it is a full self driving system. As it stands at the moment this is what their own website says:

    “Full Self-Driving Hardware on All Cars
    All Tesla vehicles produced in our factory, including Model 3, have the hardware needed for full self-driving capability at a safety level substantially greater than that of a human driver.”

    Is it any wonder drivers come to rely on the system and expect it to perform in situations like this?

    They should follow others lead, use a conventional naming system such as “Adaptive Cruise Control” and accept it for what it is. Following a standard well understood technology term will ultimately prevent drivers using the system for something it’s clearly not capable to do.

  • izzey04

    autopilot is misleading…semi auto is more apt

    • Stephen G

      no more misleading than “cruise control”

      • Vassilis

        That logic baffles me. Yes, maybe when cruise control was first introduced it was misleading but after so many years in production it isn’t. The same may happen with Autopilot but why should we repeat mistakes of the past when we have the chance not to?

        • Stephen G

          Since we understand history nobody should be surprised the name “Autopilot”, being in it’s early stages of introduction, may be misleading. Right?

2020 Hyundai Kona N With 246HP Turbo Spotted For The First Time

The go-faster version of the Hyundai Kona will likely employ the same powertrain with the i30 N.

Porsche Says It Must Look Beyond Cars To Remain Profitable

Porsche says people could buy fewer of its cars as more people live in big cities.

A BMW M2 Gran Coupe Looks Good But Is Highly Unlikely

The new BMW 2-Series Gran Coupe will most probably top out in M235i form with a turbo four.

Acura’s Iconic Integra Type R Touring Car Returns To The Track For More Action

The restored version of Acura’s Integra Type R racer can rev up to 9,000 rpm.

OMG, 2020 BMW M3 & M4 Really Do Have A Massive Grille And This Is It!

The all-new 2020 BMW M3’s nostrils might actually be bigger than the Concept 4’s.

FCA To Build New Battery Hub At Its Turin Factory

FCA’s Mirafiori plant in Turin, Italy will also be responsible for building the all-electric Fiat 500.

VW Golf Countdown: 2008-2012 Mk6 Made Safety And Comfort Breakthroughs

VW managed to shift 2.85 million units of the sixth-gen Golf, which offered a plethora of driving assist systems as well as a wide selection of petrol and diesel engines.

All-New Honda Jazz Is Hybrid-Only In Europe, Launches In Mid-2020

The 2020 Honda Jazz will pack a standard two-motor hybrid powertrain offering “EV-like driving dynamics”.

2020 Volkswagen Golf Mk8, Here’s (Likely) The First Official Picture

The official unveiling of the all-new 2020 VW Golf will take place Thursday evening on October 24.

Skoda CEO Tells Employees Not To Worry About Potential Move Downmarket

VW reportedly wants Skoda to play in the market segment of cars costing between €10,000 and €20,000.