By on October 4, 2019

The paranoid luddites that write for this site have occasionally been accused of being hyper critical of modern-day driving aids. Be it a cursory mention of how a little snow totally flummoxed the systems of an otherwise agreeable review car, the direct addressing of an issue where road salt encouraged a vehicle to attempt to steer itself into a ditch, or one of this author’s many diatribes on how the bulk of this technology doesn’t seem anywhere near market ready, there’s always a couple of exceptional individuals ready to call us backward-looking morons.

While that’s often a correct assessment in other matters, it seems we’ve called this one correctly. The American Automobile Association (AAA) recently tested four sedans from competing manufacturers, running them through a handful of scenarios intended to replicate situations that place pedestrians at extreme risk. Taking into account the above smugness, you can probably imagine how poorly it went. 

In the study, AAA used a Chevrolet Malibu (Front Pedestrian Braking), Honda Accord (Honda Sensing), Toyota Camry (Toyota Safety Sense), and Tesla Model 3 (Automatic Emergency Braking) from the 2019 model year. The testing, done in conjunction with the Automobile Club of Southern California’s Automotive Research Center, was aimed at assessing the effectiveness of the automatic emergency braking (with pedestrian detection) systems these vehicles came equipped with. Testing occurred under near-perfect conditions.

Test events included an adult crossing in front of a vehicle traveling 20 mph and 30 mph during the day and 25 mph at night, a child darting out from between two parked cars in front of a car (20 and 30 mph), a vehicle turning right onto an adjacent road with an adult crossing into traffic, and two adults standing near the side of the road with their backs to oncoming traffic (20 and 30 mph). Keen to avoid any real-world fatalities, AAA used mobile dummies as human analogs. This turned out to be an incredibly wise move.

If you wish to learn more about the finer points of the parameters and instrumentation used for testing, allow us to redirect you to the extended study. However, if you just want hear about the dummy massacre that ensued, read on.

AAA reported that the systems performed best when an adult crossed in front of a vehicle traveling 20 mph in broad daylight. Collaboratively, the systems successfully avoided a collision 40 percent of the time. But things quickly fell apart as speeds increased. At 30 mph, most failed to avoid smacking into the simulated pedestrian, and things only got worse from there.

The child-sized target darting from between two cars ended in a collision 89 percent of the time at 20 mph. The right-hand-turn test resulted in all vehicles impacting the faux humans, with only the Malibu registering that a person was even there.

When approaching the two adults standing near the road’s edge, a collision occurred 80 percent of the time at 20 mph.

Upping any vehicle’s speed to 30 mph pretty much guaranteed a collision, regardless of the scenario, and night testing proved to be the systems’ Achilles heel. None of the vehicles even registered an obstacle was present, allowing the simulated accidents to take place at full speed. That’s a bummer, especially since the vast majority of pedestrian-related accidents take place in the evening.

We hate to keep banging on the decaying corpse of this poor horse, but you’re still the best defense against an accident and should probably not rely on these (or any other “advanced driving aids”) to ensure safety. Sure, they might get you out of a jam once in a while, but you’re really gambling if you placing any amount of faith in them.

This leads us into the big question — is it responsible for manufacturers to market these admittedly flawed products? We’ve long harbored concerns that these features lull many drivers into a false sense of security, possibly encouraging inattentiveness, and there are a few studies to back that up. However, if someone’s doing their job as a motorist, then this really should be a non-issue… right?

Late in the study, AAA mentioned that the owner’s manual recovered from the glovebox of each test vehicle included an acknowledgment that “the integrated pedestrian detection system may not discern pedestrians at night or in adverse weather such as rain, snow, sleet or fog.” Automakers have also gotten into the habit of reminding drivers to keep their hands on the wheel in press releases, regardless of how advanced they claim their driving assistance or safety features happen to be. But how often is your average motorist really looking at their car’s manual? And who, other than the people that write or read about automobiles all day, have any awareness of these systems’ shortcomings?

I’ve asked around. Regular people think these features are bulletproof until they’ve been shown otherwise. That might not be a big deal when someone’s adaptive cruise control goes haywire out on a lightly trafficked highway. But who do you think the law is going to blame when automatic emergency braking fails to see that kid darting from between two cars? Until the rules change, it’s not going to be the manufacturer — which is why we should all press OEMs for better quality or, at the very least, more transparency in what these systems are actually capable of.

This isn’t a “won’t someone please think of the children” situation, and I’ll go on record as saying all pedestrians should stay the hell away from my car lest they be crushed. But the results of this AAA study are truly abysmal. If automakers are providing pedestrian detection systems that are only slightly better than nothing, why even bother?

[Images: AAA]

Get the latest TTAC e-Newsletter!


51 Comments on “Reportedly Terrible: AAA Tests Pedestrian Detection Systems...”

  • avatar

    Yikes. I at least thought the model 3 was better than this. I wonder how much difference a dummy vs a human makes.

    Definitely a good thing to know, though. The only safety feature for anything that really matters is between your ears.

    • 0 avatar

      “I at least thought the model 3 was better than this.”

      I’m wondering why you thought that? There is nothing about electric propulsion that makes Tesla inherently better at “self-driving”. People have already lost their lives due to Musk’s hubris.

      • 0 avatar

        “People have already lost their lives due to Musk’s hubris.”

        Vastly more have been saved. But the system working and avoiding an accident isn’t newsworthy. Of course you took that into account when forming your opinion, right?

    • 0 avatar

      A few years ago a popular Russian publication AutoReview tested pedestrian detection of Volvo, Mercedes and Tesla. Tesla was the only one that stopped. They were using real people, though.

      I wonder if AAA at least heated the “mobile dummies,” whatever they are, to 98 degrees?

      • 0 avatar

        Funny, but that crossed my mind while reading this. Do the cars sense heat, as well as an object?

        • 0 avatar

          How valid would that be? Cold night, peds wearing insulated snow jackets, pulled down wool caps, gloves — how much heat would actually register? Don’t pretend the clothes would be warm, assume someone just put them on and walked out of an apartment.

  • avatar

    Autonomous driving is just the latest wall street catchphrase. Many of those who invest in it don’t realize the current limitations of the technology. Mobility is another obtuse phrase wall street uses to entice investors desperate to be part of the latest fad. They will all eventually fade away like all those companies from the mid-90s.

    • 0 avatar

      This. The only good thing about these systems is that they’re the new marketing darlings for top-trim levels, so every option and feature that *actually* matters in a car is now being offered on the mid-trim cars.

    • 0 avatar

      Agreed. The task is far more complicated than proponents will acknowledge. Autonomous vehicles will work well on dedicated roadways built to communicate with autonomous vehicles. Shared roads will have to be built or modified to act as shared roads, with communication and traffic separation features designed specifically for autonomous vehicles. The trucking industry could make good use of separate roadways or lanes with autonomous tech built in.

      If you have to keep your hands on the steering wheel, eyes on the road, and stay alert and attentive at all times… you are not in an autonomous vehicle. You are driving a car that has automated assist features. Some of these are great, like ABS braking and emergency collision avoidance braking. Some of the rest of it is a gimmick, a pain in the ass, or a distraction.

  • avatar
    R Henry

    A great reality check. The technology just isn’t there.

  • avatar

    So where will this go now that there is definitive proof that these new “safety systems” don’t work? Should consumers be paying more for a car with this technology when it doesn’t work? Will people start suing the car companies when they actually run over a live person because the technology didn’t work? Or should these systems be removed from the cars until they are perfected? And how did they pass research and development procedures in the first place and get installed for us to buy?
    …. I say they should be removed

  • avatar

    I think I(we) told ya so is appropriate here.

  • avatar
    Sam Hall

    The problem with “driver aids” is more conceptual than technical. Does your lane keeping assist work all the time? Just sometimes? When does it work/not work? It quickly becomes more effort for the driver to know when to trust a given driver aid than it would be to just live without it. And at that point, having the driver aids installed in lots of cars (but not all of them) is worse than not having them at all, because in addition to not really being able to say e.g. “driver doesn’t need to pay attention from minute 2 through 8 of this ten minute drive” the presence of such systems will encourage some drivers to blindly trust them no matter what, like the regular trickle of Tesla drivers caught napping behind the wheel.

    • 0 avatar

      Totally agree Sam. Drivers who are responsible do not need these and others will become more dependent and even less responsible than they already are. I even believe the big infotainment screens need to be gone, just another distraction in my opinion.

    • 0 avatar

      My LKA doesn’t work at all. Because I turned it off. I figure if I can’t at least keep my car in it’s lane, perhaps I should stay home.

  • avatar

    Without separate and dedicated lanes for Disneycars, only one response to debates about “driver aids”…..Na, und?

    Might as well discuss what kind of cheese the moon is made of.

  • avatar

    “The paranoid luddites that write for this site”

    The cranky old man shtick is very tiring. Maybe switch it up every once and a while?

    • 0 avatar

      I think it came about from articles not totally feeling the “self-driving future” getting them constantly tarred as Luddites by the True Believers.

      Perhaps they’ll let off if the True Believers settle down. I wouldn’t count on that.

  • avatar

    I have two 2019 vehicles, both with the full suite of active safety aids. To the extent possible, I have disabled certain functions. The lane keep assist is mostly garbage. Unless its lane centering, which some cars have, it is essentially bumper bowling with the car careening from one extreme of the lane to the next and it usually only engages after you have crossed the line, in which case an inattentive driver may have already had a collision with an oncoming car. Automatic braking goes off in unexpected scenarios and scares the hell out of you. Radar controlled cruise is utterly useless in any sort of traffic as other motorists will cut you off all day because you are travelling a football field’s length behind the next car.

    Blind spot monitoring and rear cross traffic alert though is very useful and effective. So kudos for that.

    • 0 avatar

      which vehicles?

    • 0 avatar

      Good info, sounds very true as I would expect but have not experienced and would rather have a choice not to pay for these unhelpful systems on my next purchase. Make them an option for those who want this technology and save me some money so I don’t have to get them on my car.

    • 0 avatar

      I love lane keep assist for alerting me that I’m about to run off the road while futzing with a Cheetos bag when out in the absolute sticks. Ditto radar cruise when stuck behind yahoos who can’t hold a steady speed, slowing down uphill and speeding up downhill. But again, only in the hinterlands. I agree the systems are waaaay too primitive to work in anything resembling heavy traffic.

  • avatar

    “I’ve asked around.”

    I’d love to see a study on how ubiquitous the perception upon which your entire article is based really is….

    I’d bet the percentage of people who thought these systems were “bulletproof” is minuscule…

  • avatar

    Ugh, reading some of these replies hurts.

    Pedestrian Detection systems aren’t going to stop a car when a child runs out in front of a vehicle that is doing 30mph. Most drivers wouldn’t be able to stop in time for these tests either. Manuals of cars equipped with these technologies are very succinct about where they work and that they are no substitute for the driver.

    Pedestrian Detection systems work wonderfully below 20mph, in urban environments, and when pulling out from a parking garage or around a blind corner.

    They are “driver assist” technologies, not “driver replacement” technologies, and in that context they work very well (and have saved many from accidents).

    • 0 avatar


      Look, nobody should expect the driver safety aids to work 100% of the time or prevent everything.

      But if it works even 10% of the time, that’s not just a statistic. That 10% represents actual people who didn’t get hurt. Actual accidents that were avoided. And there is zero downside. Having backup doesn’t make you a worse driver unless you were incredibly shitty to begin with.

  • avatar

    With just a few million autonomous vehicles on the road our highways would resemble the film Death Race 2000. Still, I am now not worried because this fad is going nowhere fast.

  • avatar

    The system in our 2012 Volvo once confused a curve in the road for me about to hit a pedestrian on the sidewalk. The pedestrian was never in any danger. But my underwear certainly was, thanks to the unexpected sudden application of the brakes, the loud buzzer, the flashing red hologram, and the big black F-150 that nearly rear-ended me.

  • avatar

    Why bother?

    Because slightly better than nothing IS STILL BETTER THAN NOTHING.

    • 0 avatar

      Not necessarily, because it doesn’t cost nothing. If it costs $1 million for every life saved, is it worth it? Probably, though a lot of “low value” lives could be saved for that much instead. What about $10 million?

      Probably not a big concern, as eventually this stuff should end up being quite cheap. They just need to make it reliable. If it activates unnecessarily like my girlfriend’s 2019 CRV did at highway speed during our last trip, I’m disabling it. Even worse would be if it wouldn’t allow you to escape by intentionally running over somebody who was threatening your life.

      • 0 avatar

        Components,sensors,even processors should not be terribly expensive. It’s the sheer volume of variables associated with a road-going vehicle that presents the problem. I’m firmly convinced that it’s far easier to program for flight than it is for automobiles. Also, in a lifetime of dealing with digital demons, I have yet to meet a processor incapable of dropping a bit now and then. So there’s that. Far better that we should drive than further enable the lack of attention becoming more prevalent on our roadways.

      • 0 avatar

        Except it’s not 1 million. You can get some sort of driver aid on most cars for a few grand. And it’s entirely optional.

        My objection is when people say that because it doesnt work perfectly yet, nobody should be allowed to have it at all. That’s nonsense.

        • 0 avatar

          A few grand times the number of vehicles with it, divided by the lives saved.

          But the pre-collision braking system might pay for itself by reducing other collisions too. Insurance rates should tell the tale.

          I agree that the tone of the article is overly cynical.

          I’m fine with anything that the consumer has a choice in purchasing, and allows full user control of the system.

    • 0 avatar

      In my experience nothing can be better than something. I was driving a friends Jeep Grand Cherokee through a winding mountain road last winter. Road was snow covered in most places, and it’s single lane in each direction with no shoulder. Temperature was just below freezing, but in a few places there were gaps in the trees where the sun shone on the road and melted the snow down to pavement, exposing the yellow line.

      Passing traffic has caused the snow to pile along the center line where the road is in the shadow of the trees. That is loose snow and than can tend to pull a car into it, so I was staying toward the right side of the road.

      All is fine until I came to a bigger gap in the trees, probably a car and half long where there is enough melting to expose the centre line and the road is a mix of water and ice. Now the lane keeping has enough line to see and it determined that I was out of my lane and corrected hard enough to pull the wheel out of my hands and lurched the car to the side to attempt to bring it back to the center of the road.

      The problem was that by the time it did this I was past the sunny spot and back into the snow covered road, so the upsetting of the car by the sudden steering input put the car into a spin and I ended up doing a 180. My only saving grace was there was no other traffic, as I would have hit anything that was in the other lane. The other good thing was I managed to stay on the road by pure luck, and didn’t end up in the ditch or off an embankment.

      The last thing I need is the computers onboard causing their own accidents by “helping me” in the same way as the 737 Max.

      Everyone who knows how to drive in winter, knows you drive smoothly and gently, with no sudden inputs. Apparently the computers are not able to cope with winter.

  • avatar

    I write this in the sad hope that these “safety” gimmicks will become a bad memory – or at least be made optional.

    I recently dumped a new Mazda 3 because its silly active braking system mistook an expansion joint for an obstruction. When your car slams the brakes on for no reason on the highway, it gets your attention. Sadly, I now drive one of the most basic new cars out there since it’s one of the few new cars without “active braking”.

    Be advised – I have not found any manufacturer that allows the driver to fully disable systems such as active braking. Even if you’re allowed to “disable” active braking (usually only on a per-drive basis), the system is never really off and will still step in if it thinks there’s a problem.

    Call me a Luddite, but this active safety stuff is damn stupid…

    • 0 avatar

      Totally agree Goose.

    • 0 avatar

      I believe I can disable the collision warning/emergency braking on my Toyota, but its the only one I leave on because it generally doesn’t cause a problem and it might come in handy (hasn’t yet really but you never know). I disabled the lane departure and whatever the third system was because they were too annoying.

    • 0 avatar

      “ I recently dumped a new Mazda 3 because its silly active braking system mistook an expansion joint for an obstruction.”

      Sure you did…

      • 0 avatar

        Quite true, actually. Here’s the listing for the car I dumped. It appears that the price has dropped below the amount they gave me in trade value. I’m sure the dealer would be happy to tell of the guy that traded-in the almost-new car:

    • 0 avatar

      Shame to hear about your 3. On the ’14, active braking was part of the ‘Tech Package’, and the only thing I wanted out of that was the radiator shutters…so no tech for me. The 2.0/6M is not extraordinarily fast, but it is extremely quick, living for corners…especially after dumping the terrible OEM Yokos in favor of something that actually provides a bit of grip. Three and a half years on the Yoko LRRs at 41.9 mpg, 2 so far on the stickies at 40.9. I’ll take that trade.

  • avatar

    I have to agree Goose.

  • avatar

    The lane-keep in my new Accord has centering capability, and does take a little bit of the edge off so you can concentrate a little more on situational awareness. I think I’ve had a couple instances where I caught a potential hazard a half-second sooner because the car was actually following a slight curve. The system is very good about warning the driver if they don’t have a hand on the wheel; if you ignore the warning, it will disengage. Additionally, the system must be engaged with each ignition cycle, and can only be used if the ACC is also on, regardless if ACC is being used.

    This is my second Honda (and car) with adaptive cruise, and the implementation in the 10th-Gen Accord is much smoother and less abrupt; the full-stop capability is damn near flawless in operation!

    I may have had one auto-brake application, in a construction zone after I had already slowed appropriately — it felt like the car kept pressure on the brake for not even a second, and all that resulted was that I ended up a little further behind the preceding car; I had plenty of following distance, in any case. Other than that, no false alarms whatsoever! I’ve had the car since June 6th, for 3,100 miles.

    No other complaints: my blind-spot system has maybe caught one car that I wouldn’t have seen in the mirror, but would have caught on a shoulder check, and the rear cross-traffic monitor has beeped immediately upon selecting Reverse, before I could turn my head to look, or even glance at the camera! But I’m preaching to the B&B choir when I say that I won’t solely rely on any of these ADAS things for anything! Perhaps the only thing I’ve come to rely on is the backup camera, as I get a better view behind the car without having to turn my head, at least in a straight line, as if I’m backing out of my garage onto my condo’s parking apron. But you’d better believe I’m going to glance over my shoulder occasionally if I have to turn in any direction while reversing, and in a parking lot or something similar, I’m going to go between glancing over my shoulder and checking the camera, and the cross-traffic alert is a backup!

  • avatar
    AccordStu did a test of pedestrian detection systems and found that nine out of the eleven tested slowed the cars substantially when collision with a pedestrian was imminent. That is one of the reasons I bought a 2018 Accord. I drive as if the car has no driver aids, like I have for the past fifty years, but I figure that these systems could one day be the difference between killing and not killing someone.

  • avatar

    On my 2019 car, I find only the radar ACC and rear cross traffic alert work as I would want them to. I’ve experienced one sudden stop where my easing off a sudden brake application when the situation changed apparently meant to some defective software that I really wanted to stop regardless. Nope. That braking when not needed in many differing situations seems to be the major fault of all these system across vehicle brands. In particular, stopping dead regardless is not a fail-safe default mode to a proper logician.

    The RCTA actually picks out humans walking off-camera from the sides in crowded parking lots when backing up. Nice. When exiting facing forward with pedestrians about, however, it takes a nap even when the trolley hut or a RAM obscures the view in one or both directions. So you cannot rely on this crap, and this forward error fits in with the article’s story that pedestrians are not well identified. You got me if you wonder why it works in reverse. Haven’t a clue.

    LKAS is overall useless, let’s get that one clear. At least my vehicle has a physical on-off button for it. So I just have to bear a constant light on the dash in exchange for no ghosts in the steering.

    The BLIS mainly goes off when I’m in two lanes of traffic which are turning side-by-side at once, and ONLY during the turn – some bright spark must have decided steering wheel turning should be the reason to blare a useless warning and flash lights in my rearview mirrors (which aren’t exactly what I’m monitoring in that situation). Oh, it’s roundabout heaven! BLIS’s other major concern seems to be when pulling back in after passing on a two-lane road. Only a mile of free space behind satisfies it by not setting off klaxons.

    Overall, it’s rank amateur BS. Although I have to give an Acura TLX I test drove the Blue Ribbon for being scared of its surroundings. A mile in urban conditions of that nonsense, and I turned around and drove back to the dealer, the salesman nonplussed. No thanks.

    Automating this unfit-for-service stuff and declaring it good for transporting people automatically is the height of folly. A lot more work needs to be done first, perhaps beginning by employing programmers who actually drive cars on a daily basis, and not some young whizz-bang bicycling nerd.

  • avatar

    it’s interesting that the Chevy consistently saw pedestrians much earlier than the others, but took no action. Do they consider not braking to be conservative? They clearly have the tech to be doing something earlier.

  • avatar

    Run a stupid test, get stupid results. You don’t have ROOM to brake from more than 20 mph before you’re going to hit the person/dummy regardless in most of these scenarios. That’s why Volvo calls theirs City Safety: it’s designed to work in city conditions i.e. under 20 mph.

  • avatar
    Jeff Semenak

    Car & Driver article from last November.

Read all comments

Back to TopLeave a Reply

You must be logged in to post a comment.

Recent Comments

  • GenesisCoupe380GT: I’m grateful Cad has this option at all. BMW took it away from the M5 and now it’s...
  • ToolGuy: That hood graphic reminds me of the circuit board out of my WiFi router which suffered a lightning strike in...
  • ToolGuy: Those A-pillars look refreshingly slender (could be the camera angle).
  • EBFlex: Built Ford Proud! This is a result of the furniture salesman squeezing suppliers to the bone to keep the...
  • ToolGuy: The new Corvette has a workaround – pop the front trunk. (It must function as a fighter-jet-style...

New Car Research

Get a Free Dealer Quote

Who We Are

  • Matthew Guy
  • Timothy Cain
  • Adam Tonge
  • Bozi Tatarevic
  • Chris Tonn
  • Corey Lewis
  • Mark Baruth
  • Ronnie Schreiber