Driverless vehicles commonly used on public roads are still years in the future. There are no meaningful stats on the safety of such vehicles since they aren’t really in wide use yet.

No, mile for mile, driverless vehicles were more than 26 times as lethal as piloted vehicles:

https://www.eschatonblog.com/2018/03/the-faith-is-strong.html

And that's all piloted vehicles in all kinds of conditions across the country. Driverless vehicles were operating under far more optimal conditions.

You'll see lots of anti-Tesla articles posted on his site.

I think current automobile sensor technologies are pretty good driver's aids, but of course there's still a ways to go before fully autonomous road travel arrives. But that's where things are heading; just look at the vehicle-aware technologies car companies are touting in their current models.

And of course sensors fail. The way to avoid the effects of a sensor failure is to add redundancy into the design.

Atrios has his reasons for disliking Musk and Musk's boundless affinity for making promises he can't deliver on. He has been quite good at documenting those promises and Musk's failures to deliver. Atrios has also made the point quite often, and he's quite right, that there is a huge difference between 99% autonomous and true autonomy. If you have to have a driver at the wheel at all times ready to take over when the car is about to get into a jam (and not when it's too late, hopefully), why not just have a driver?

Atrios also is at pains to point out that practically all autonomous driving logged so far has been under really good conditions, clear roads and good weather. Will autonomous vehicles perform flawlessly in inclement weather like ice and snow? A traffic accident blocking 2½ out of 3 lanes? Coming upon a stalled vehicle in the middle of a lane? A crowded parking lot during the Christmas shopping rush?

I'd bet I will not live to see a truly autonomous car (I'm 60), and I'd bet you won't, either.

It was telling me that my antilock brake system had gone up, and as a result, it wouldn't let me get my car out of park.

Of course after inspection there was nothing wrong with my antilock brake system. It was only a problem with a sensor to the antilock brake system.

It was rather annoying to say the least.

and would not let you drive the vehicle. As it turned out, the only problem with your brakes was a failed sensor. You could have driven it safely, since your ABS system is a relatively new thing. You drove cars before it existed safely. But, the computer was programmed to not let the car move if that sensor failed. It was over-cautious.

Here's what I wonder: what would the car have done if the sensor failed while you were driving? Would it have applied the brakes automatically and forced you to stop driving? The answer is that it would not. A light would have come on on your dashboard, and it would be your decision about what to do. The computer is also programmed to deal with that issue.

However, it failed safe when your car was not moving. The reality, however, is that the ABS system is not an essential system. You can still drive safely, even if that automatic system is non-functional. You might drive a bit more carefully, but you could have driven to your dealer or auto repair place and had them check it. But, whoever programmed that computer decided that you should not drive at all if the ABS system was non-functional. You lost control of your vehicle until it was fixed.

to escape a forest fire or an incoming tsunami.

Having a car that won't go out of Park when you're sitting in the driveway, though, is failing safe. It's annoying and unnecessary, though. A working ABS system is not an essential safety system. It is enough for it to display a warning light to let you know that it's off. On my car, there's a button I can push to turn it off. If I do, that warning light goes on, but I can still drive. I haven't shut it off, so far, because I haven't been in a situation where it needed to be shut off, but I'm glad I have that option. I do sometimes shut of the automatic stability control system during the winter, since it sometimes makes driving on snow and ice less safe, in my opinion.

I don't mind such systems, as long as I'm in control of them when I need to be.

By flying at night, I mean over North Korea or the desert, where there are no city lights to act as a guide. I also exclude astronomers as pilots, because they would know what stars are in front of them and thus whether they were flying level or not.

They also have artificial horizon sensors and instruments. However, planes only recently have autopilot systems that react automatically unless overriden. In the case of that Boeing 737 Max 8, they can react so dramatically that the plane flies itself into the ground and kills everyone on-board.

In the case of those two crashes, the pilots, apparently, had not been trained on what to do if what occurred occurs. It's fairly simple, really. The pilot or co-pilot disables the automatic control system and flies the plane him or herself. But, such an alarming change of attitude in flight could easily disorient a pilot, who might try to override the automatic control system without disabling it. That is apparently what happened, and it's likely that training was inadequate.

Fully functional sensors, like angle of attack sensors, are important, for the reasons you stated. Flying "by the seat of the pants," can easily lead to disaster. However, no automatic control system should do what happened on those two flights. Either through redundancy of sensors or programming that does not allow unlimited control to the automated system, the system should have had limits as to what control it could have. No pilot would overcompensate to the extent that the automated system did, if other signs of an improper angle of attack were not present.

The human brain is superior to digital controls in that it allows incremental judgments to take over. If the "angle of attack" instrument shows a reading that is inconsistent with the perceived angle of attack by the pilot, the pilot can reason that it might be the sensor or display that is faulty. The pilot would not suddenly put the plane into a steep dive, as the automation did.

Have multiple instruments that provide overlapping/redundant information. By crosschecking (a skill taught from the outset of flight training) various instruments, you interpret and varify not only the attitude of the plane, but the accuracy of the various instruments.

As example, your artificial horizon shows a climb. Check you airspeed, altimeter, vertical speed, and rpm. See if these agree with the artificial horizon.

I find it amazing and hard to fathom that the computer system in the 8s would not do a similar check.

JW

on the road, human operators have demonstrated a very long history of causing crashes through their own problems such as impaired driving, which is the organic equivalent of a faulty sensor or failing cpu.

driverless cars really only need to be safer than their human-operated counterparts. sadly, we set a low bar.

what's we find disturbing about driverless car accidents is that we have come to understand and accept human error. driverless car errors are foreign concepts to us and seem like a new and unnecessary risk.

but in fact, it's an alternative risk as driverless cars eliminate the risks of drunk driving, sleepy driving, texting while driving, etc. they just replace it with other risks, such as faulty sensor risk.

once we get over the different kinds of causes or crashes, they start to make sense once they're merely safer than human driving.


personally i think the biggest problems are the mix of driverless cars and old-fashioned human-operated cars. the roads will probably be at their safest when *all* vehicles are driverless and then one vehicle with an issue may be able to cause all nearby vehicles to slow down or stop, e.g.

of course, the standard airplane design philosophy of redundant systems would also be a good idea for driverless cars....

people are pushed to buy more of these devices and gadgets for their homes. When your appliance stops and suddenly its a bad sensor. People have blind faith in these things. Human error is now sensor error.

Cars are already on the road that depend on sensors and computer software. Pretty much all of them in fact.

My F150 pickup for instance. Has sensors that detect if I'm approaching a car to rapidly and likely to hit it. What does it do? It automatically applies the brakes. If I start drift off the side of the road, or into the other lane, it will move the wheel back. It also speeds up and slows down automatically based on the speed of the car in front (using brakes or gas)

That's just a pickup truck. All the technology is on a 3 year old, non luxury vehicle. I'm sure that new vehicles, especially expensive ones have even more.

What's to stop my pickup truck from slamming on the brakes on the highway in front of school bus? Or turning the wheel into the bus or the tanker truck next to me? A: Software and sensors.

I hear ya man. But that cat is out of the bag on this.

California. It had all sorts of sensors, designed to alert me of dangers. One of them was designed to alert me if a car was in my blind spot. I had to shut that feature off, though, because it was constantly alerting me. I was driving on I-405 from the airport. There was ALWAYS a car in that blind spot, and on both sides of my car. So, the alert was constantly going off and was very distracting.

So, I shut the alert system off and used my experience in driving on that horrible excuse for a highway instead. That way, I wasn't being distracted by a safety system that was functioning too well for my safety.

We'll be buying a new car later this year. It will, no doubt have many such sensors. I'm looking forward very much to the back-up camera system. I think that's a wonderful thing. I'll probably enable some of the other systems as well, when they are appropriate for a particular situation.

I won't let automation drive the car, however, ever. I will drive the car. I will decide what to do, based on my experience, and my own sensors and decision-making abilities. I'll give self-driving cars the same wide berth I currently give to trucks transporting liquid oxygen.

When these sensors fail, bad things can happen. But it's a risk we are willing to take because of the convenience.

All gas furnaces produced today shut off automatically if any of the safety sensors fails to work. The digital control board shuts off the gas valve and the furnace immediately if the sensor does not provide the expected data. It's a very simple program. If the sensor's output is not in the acceptable range, the furnace shuts down.

Now, that can be inconvenient, because the house starts getting cold, but that generates a call to an HVAC person who can diagnose and repair the problem.

Shutting off the furnace is the appropriate response. There are even safeguards included to prevent the furnace owner from bypassing those sensors. Before the furnace starts, the state of each sensor is measured. If it is bypassed by the owner, the furnace will not start. Modern gas furnaces do a good job of this. Older ones, not quite as good.

Fail-safe automation is not really a major problem. That is the system used in heating systems. If there is a flaw, the system will simply not operate at all, forcing the owner to get the thing fixed.

.... to prevent really bad things from happening, that fails and prevents your furnace from operating. Because they are delicate electronics susceptible to wear.


I’ve flown planes with stick shakers and stick pushers. What I don’t get is the pilots in the MAX can’t overpower the system.

I guess some systems don’t allow an override by control inputs and rely solely on the pilots knowing when and how to QUICKLY disable the system versus ignoring/overpowering the pusher. Tricky at lower altitudes.

There have been several accidents where pilots have missed messages on the screen (crew alert system) that tells the crew what the plane is up to. Apparently easier to do than one would think when bells and alarms are sounding.

Perhaps the solution is as simple as a more prominent disable/override system. Not to mention better mandated training.

No automated system should override pilot input like that. Apparently, the knowledge of how to disable the system wasn't immediately available to those pilots. I have no idea what the cockpit looks like in that plane, nor where the disabling switch might be. You'd think a system that is capable of overriding pilot input would have a bright flashing line on the disabling switch when it was working.

Weird incidents. Scary.

they rely on work. I make it a point to learn about systems in my home. HVAC systems, I understand because I've written dozens and dozens of pages about them, so I've learned. I finally had my 17 year old furnace replaced a couple of years ago. I had fixed it myself a couple of times after diagnosing which component had failed and stopped it from working. The first time, it was the draft inducer fan which had a bearing freeze up. The next time it was the vacuum switch that checked that fan. Parts are available online and it's not rocket science to work on a furnace.

Installing a replacement furnace, though, involves skills I don't have, like sheet metal and duct work. I could do the rest, but why? When the new one got installed, the installer left the installation manual for it, which included a complete operational system diagram that showed the control board logic. As I always do, I read the whole thing and have it stored near the furnace.

I've read installation manuals and service manuals for maybe 2 dozen furnace models so far in preparation for writing stuff for a client. I tend to over-prepare for writing assignments.

It’s the draft inducer motor. $52 bucks for the motor, $15 bucks for the fan wheel, and $7 bucks for the gasket. Parts will probably be here Monday.

2 pieces of pizza for the service call next door last night. And a home cooked dinner at a later date for the install.

I like being self sufficient. Not only do you save on the labor but waiting around for the service call is a huge waste of time. Often times twice for the return trip after the parts are ordered.

4 hours times 2 waiting around when I can either go pick up a part or order online if not in a rush? Works for me!

At 73, I'm just not in the mood somehow. It's nothing to replace a draft inducer, though. The problem is getting the parts. Here, the local HVAC supplier won't sell them to you unless you're in the business. So, you have to order them online. It's easy enough to find parts numbers, it you know what to look for, but you're going to wait for the parts. That's not an option in Minnesota in January, when the inside temp is dropping 2 degrees an hour. So you call the HVAC guy, and he charged you away too much.

And there it is, except the furnace guy wants to sell you a new furnace...And so on.

If the flame sensor for instance fails, the gas valve closes. Valve solenoids are designed to pull against gravity or a spring, if the circuit fails, the valve shuts.

Everything that performs some function uses some kind of sensor. Just name a system, appliance, machine, or animal and it will have sensors (outside of simple mechanical machines).

Temp sensors, speed sensors, voltage, current, direction, etc, etc, etc....

Don't be a ludite.

What I am, though, is aware that the human decision-making process is more flexible than a computer-based one. It is that inflexibility and the limitations of a hard-coded response to inputs that is the problem, actually, in situations where there are so many variables, like driving a car or flying an airplane.

The intuitive responses of humans, assuming that they are paying attention, are better equipped to solve analog problems than are computers. And everything about driving a car or flying a plane involves analog calculations. Humans learn the skills of such operations over time, based on exposure to a wide range of situations and conditions. If they're paying attention to what they're doing, they are likely to do the right thing to avoid an accident, almost automatically.

The answer, then, is for humans to pay better attention, so they can have the kind of accident-free history over almost 60 years that I have. I have several million miles of road time driving a vehicle. I'm pretty good at it by now.

I'm not a luddite. I'm an experienced driver.

But we aren't faster, or more durable, or more precise. We don't have the endurance or the brain power to keep up. It's just reality.

It's fine to say people need to be more aware, but it isn't practical. In the case of autopilots specifically, they have been shown to cut down on pilot/captain fatigue. They allow you to pay attention to other important functions while the AP does the mundane.

Humans aren't better equipped. That is a fallacy imho.

This comes from someone who uses and relies on these systems. I have a rather slow fishing boat. When I go out 100 miles, it's an 8-10 hour trip each way, with hundreds of course corrections an hour to maintain heading. That can be downright exhausting. Also, in cases where there is no visibility, it is very easy to go in circles while distracted looking at radar or charts. The AP holds a steady course, steadier than I can do by hand, and I can concentrate on not hitting anything.

The AP in my case actually does a better job of holding a course than I can.

I can give you 1000 other examples of process control systems where humans would be absolute failures. Things would blow up. Automated systems work.

Can complexity and poor engineering create risks? Sure, but so can a homer simpson at the controls of a complex system.

I put a lot of faith in automated systems, but also use a lot of caution.

Don't be a ludite. People can't do a better job in a vast array of applications. We are too slow, too fragile, and too stupid in a lot of cases.

I hate people, but I respect them for what they are good at. People aren't unecessary by any means. Just use them for what they are good for. Allow them to step in when it's clear the system isn't working. But let the systems do what they are good at.

to discuss. Misanthropy is useless in a world populated with over 7 billion people. People create and design every system, so you're dependent on people in one way or another in everything you do.

Annual Global Road Crash Statistics:
Nearly 1.25 million people die in road crashes each year, on average 3,287 deaths a day. An additional 20-50 million are injured or disabled.

I learned to drive in a stick shift VWbug as did three of my siblings. Mom actually taught dad to drive.
And I love driving and don't want any automatic car near me.

My 6-speed KIA Soul is a cool vehicle. However, I don't mind automatic transmissions, either. My next car will have one, as have a number of my previous ones. How a car shifts isn't really of any concern to me. It's the nut behind the wheel that concerns me, whether it's a human being or a computer.

I, too, take operating a 2000-pound death trap very seriously but kinda think some people may take it too casually. Smartphones may be causing some chronic distraction symptoms in some people even when they aren't using them.

I heard somebody say the fix for the sensor problem may involve using two sensors instead of one. I suppose the two would have to agree before taking any action.

that's just another decision for the system to make.

which is not uncommon at all in many systems. Then, pilots also need a visual and/or audio alert that there is a disagree in the first place. Helps them troubleshoot if it becomes necessary.

If two are used and disagree...which one is correct (especially flying in the clouds or at night)?

It's all about odds. What's the likelihood that at the same time, two are bad and one is correct - versus two sensors matching and the third one is not.

Secondly with the Boeing MCAS system we are all talking about: Currently, only one sensor is used. In avation safey that is Pathetic...

But, ya know, doing that might take a big chunk out of the CEO's multimillion dollar bonus this year so...maybe next year, K? Without all those millions in compensation they'll never find anybody to sit around in the corner office pretending to be superhuman, ya know.

MCAS was only connected to one of them.

Safer, more environmentally friendly, more cost efficient.

The highest price we pay for car crashes is in the loss of human lives, however society also bears the brunt of the many costs associated with motor vehicle accidents. According to the National Highway Traffic Safety Administration (NHTSA), U.S. motor vehicle crashes in 2010 cost almost $1 trillion in loss of productivity and loss of life. The study was released in May 2014. The auto industry's steady improvements in vehicle safety over the last several decades — despite a litany of safety recalls — had driven down the number of roadway deaths to an all-time low of 32,675 in 2014.

NHTSA reports the number of people killed on the road in the U.S. soared 7.2% to 35,092 in 2015, marking the deadliest year on the road since 2008. Though the increase was widely expected after NHTSA last month revealed a preliminary estimate of a 7.7% increase, the official figure solidifies 2015's dubious distinction as the first year-over-year increase since 2012. In addition, roadway deaths of pedestrians and cyclists hit a two-decade high in 2015.

New findings from the Insurance Research Council's (IRC) Auto Injury Insurance Claims Study shows that medical expenses reported by auto injury claimants continue to increase faster than the rate of inflation, in spite of the fact that the severity of the injuries themselves remain on a downward trend. From 2007 to 2012, average claimed economic losses (which include expenses for medical care, lost wages and other out-of-pocket expenditures) grew 8 percent annualized among personal injury protection (PIP) claimants. Among bodily injury (BI) claimants, average claimed losses grew 4 percent. Over the same period, measures such as the percentage of claimants who had no visible injuries at the accident scene or who had fewer than 10 days in which they were unable to perform their usual daily activities provided evidence of a continuing decline in the severity of injuries.

and safety yet. There simply aren't enough of them in operation to make any comparisons.

I'm hoping we don't ever see that data. I don't think we're going to like it very much at all.

The current stats, as highlighted above, are horrendous.

It only makes sense to look in that direction as technology advances.

I suspect the truth about autonomous vehicle safety will come out soon enough. I have no doubt we'll be encountering them on the road before long. As I said, I'll give them the same wide berth I give liquid oxygen transport trucks.

Or the drunk in the other lane. Or the person texting. Or the person eating a big mac. Or the elderly person whos vision has dramatically decreased in the last year. Or the....

It may end up being a combination of things, with the common item being pilot training.

Pilots are trained to react to instrumentation failures. The issue here seems to be that the training didn't match up with the software being used. That said, in our products we spend an inordinate amount of time designing in fault sensors to detect that something isn't working correctly. I'm a bit surprised that there isn't logic in the system to detect this failure.

Our entire optical processing apparatus is incredibly unreliable. We just discount that failure rate as baseline acceptable. Once a system is better than that I’m happy to swap humans out for it.

they're killed because of human operators.

automatic doors, remote controls, alarms, we are so far in to the electronic revolution.

An autonomous vehicle is not like any of those things. Let's take your first example:

If I walk up to a door to enter a building or leave it, and it opens automatically, then I walk through it. If it doesn't, I operate the door manually. The automatic door has just one function - to open the door when a human approaches it. Very simple system. If it's not there, then I open the door myself.

Same thing with a remote control. I push a button and the thing I want to happen happens. If I did not have the remote control, I could walk over to the device and operate it manually. Is it a beneficial device, that remote? I suppose so. If I were handicapped in some way, it might even be an essential device, but it is a simple device that merely does what I want when I press its button.

At best we might have what could be considered "remotely piloted" cars. Mostly because the legal system wants someone to blame.

The point is that the MCAS system as implemented is not reliable and fails in a very nasty way that the pilots did not understand.

A quibble: the angle of attack sensor measures the angle that the airstream passes over the wing. A plane flying straight and level at cruising speed would have a low angle of attack. Maintain the same attitude, but reduce the speed and the angle will increase as the plane settles through the air. Make the speed low enough, the AOA gets too large and the wing stalls and the plane plummets.

.....as the Air Bus stall warning deactivates when the airspeed drops below a certain limit. (I guess they figure you are on the ground or have faulty air speed indications)
So the stall warning came back on whenever the pilot pushed forward on the stick and sped the plane up and then stopped when he pulled back, this happened several times during the stall and this apparently confused the pilot.
The stall warning sounded continuously for 54 seconds though.
I still can't imagine a pilot though who thought pulling back on the stick would get them out of a stall. And the alt indicators were dropping like a rock.

"An "angle of attack" sensor is a simple device on airplanes. As the plane flies, it measures the attitude of the plane with regard to the horizontal angle at which the plane is flying along the line of flight. Such sensors have been around for a very long time, but recent automation of flight software has made them more critical. Now, if they fail, the plane's software automatically reacts. Before automated flying, pilots would see an aberrant reading on their "angle of attack" instrument, check that angle visually, and continue to fly the plane. With an automated system, that same failure can result in automatic actions which the pilot cannot override without shutting off the automated controls. Two Boeing 737 Max 8 aircraft have crashed due to this. "

The above description is not correct.

The angle of attack of an aircraft is the angle between the chord line of the aircraft's wings (and imaginary line connecting the leading and trailing edges and the direction of the air coming toward the wings (from the point of view of aerodynamics, there is no difference between assuming the air is moving and the aircraft is stationary vs assuming the aircraft is moving and the air is stationary). In cruise flight, the angle of attack is small and so is nearly the same as the aircraft's "attitude" - the angle between the aircraft's longitudinal axis and a horizontal ground plane. In most other flight conditions, the two angles can be very different and it's not possible to observe angle of attack by looking out of the window. Case-in-point, during landing, the aircraft is relatively slow and angle of attack needs to be high to maintain lift, but the aircraft's attitude angle is low.

The angle-of-attack sensor tells the pilot how close the aircraft is to its stall angle-of-attack. Contrary to popular belief, an aircraft can only experience an aerodynamic stall if the angle-of-attack exceeds a critical value, typically around 12-15 degrees for a commercial airliner. This sensor is typically linked to a device called a stick-shaker that causes the pilot's control yoke to vibrate if the critical angle is approached. In addition, there may be a "stick pusher" that will also push forward on the yoke to lower the nose and reduce angle of attack. Note that this applies to most commercial aircraft except Airbus. Airbus' planes are fly-by-wire and the flight control software normally does not let the pilots exceed the critical angle-of-attack and stall the plane. However, when sensors fail, as in the case of Air France Flight 447, some of the built-in safety is lost and in that case resulted in a fatal crash.

It is considered bad practice to allow flight control software to be tied into a single sensor. Typically, there are three of each type of sensor and the flight control computer is constantly polling them for disagreement in readings. If one sensor disagrees with the other two, it's input is discarded as unreliable.

On the 737 MAX8, Boeing added a system called MCAS (Maneuvering Characteristics Augmentation System) to the flight controls to overcome an aircraft certification problem. In simple terms, as the pilot pulls back on the yoke to raise the nose of the aircraft the force required to pull the nose higher should increase. In the 737 MAX8, this did not happen because of an aerodynamic effect caused by the size and extreme forward location of the engine nacelles. The aircraft quite simply was not able to be certificated for commercial flight by the FAA if this condition was not corrected. Boeing's solution was MCAS, which commands nose down pitch trim by the horizontal stabiliser when a certain angle of attack is reached and some other conditions are met. In Boeing's certification documentation, they indicated that the system could command a maximum of 0.6 degrees of nose down trim. This was found to be inadequate and they changed the system to be able to command 2.5 degrees of nose down trim but did not update the certification documents to reflect this.

Importantly, the MCAS system resets if the pilot attempts to change the horizontal stabiliser trim manually. Thus, in the case of a failed AOA sensor that is indicating a high angle-of-attack the system would keep commanding 2.5 degree increments of nose down trim each time the pilot would try to manually override it.

In the case of the Lion Air 737 Max, the aircraft flew at least three times before the fatal crash with a faulty sensor that triggered the MCAS to make unwanted trim inputs. Boeing never told the pilots about the system, but they were able to work around the unwanted inputs by shutting off the auto-trim system entirely and trimming the aircraft manually. It's important to note that the 737 trim wheel is nearly constantly in motion during flight so shutting off the system would increase pilot work load significantly. In addition, it would make it harder for pilots to detect unwanted trim activation. On the the flight that crashed, it seems likely that the flight crew never did figure out why the aircraft kept wanting to nose dive.

We know less about the Ethiopian crash, but it seems evidence has been found that shows the aircraft crashed in a maximum nose down trim condition.

in both cases, had the pilots known to shut of the auto-trim system it is likely that neither crash would have occurred. In fact, we know that the Lion Air aircraft flew several times with the same broken system that ultimately resulted in a crash.

There are several issues at play here:

1/ the appropriateness of Boeing's solution to resolve the certification problem with MCAS
2/ tying the MCAS system to a single AOA sensor
3/ Boeing's transparency with the FAA in not updating certification documents when the specifications of the MCAS changed
4/ Boeing's decision not to document the MCAS system in the aircraft operating documentation it issued to airlines

The proximate cause of both of these crashes is going to start with the flight crew's improper response to a control emergency, so-called "pilot error." However, the 4 issues above will determine whether it was reasonable to expect them to respond correctly in their particular circumstance.

None of this makes Boeing look particularly good. However, this is not an indictment of automation per se. As aircraft have become increasingly automated, they have undoubtedly become safer. The last major accident by a US carrier was American Airlines Flight 587 in 2001, nearly 18 years ago. You can't argue with that record of safety, and as much as it is due to better training, procedures, and design, aircraft automation is also a significant contributor.

Does that mean I'm ready to get on a fully automated commercial aircraft? Not today, but the day is coming.

Engineering speak for “FUBAR.”

It appears that the reason for the fatal crashes had as much to do with pilot training as it did with flaws in the MCAS system and the plane's design issues. The use of a single sensor, as well, is a serious flaw.

Unfortunately pilot training varies, as do the skills and training retention of individual pilots. The combination of factors led to a couple of disastrous crashes that should not have occurred.

Where does the blame fall? I don't know exactly, but the fix will have to be better performance in all areas, I think. What will it take to get the 737 Max aircraft back in service? I suspect some software change and some additional training materials and training sessions will be the answer to that. Will that be enough? Possibly. It's too late to make design changes that will affect aircraft already in service, especially the factors that cause the need for the MCAS system. So, the response will be patches and work-arounds. Too bad, really.

There are serious issues with FAA's oversight of the aircraft industry, I think, along with some conflicts of interest that have financial roots. Air travel is a marginal industry financially. That has led to some compromises, despite the excellent safety record.

It is interesting that none of these crashes have happened with any US airline. Perhaps that should be looked into. I suspect more care is given to training and training updates here.

....it is not clear the pilot knew the plane could indeed be flown into a stall when the computer switched to alternate law due to faulty indicators. (frozen over pitot tubes)

Air travel is safer now than EVER! 10 years in USA and one fatility.
Air travel was risky in the 1930s, etc but TECHNOLOGY improved and it will improve with self driving cars also.

Wow, you obviously don’t realize 30,000 people die a year due to driver error.

Read more!

Asiana Airlines Flight 214
Atlas Air Flight 3591
Southwest Airlines Flight 1380


The big thing is you compare a relatively small number of trips with large passenger counts to a far larger number of trips with as few as one passenger. 43,000 flights per day in airplanes vs 1.1 Billion trips per day in cars. Yes, the body total is bigger for cars, but the number of trips is orders of magnitude larger with far less restriction in operation. The push for automated cars would significantly reduce those trips, a factor that I expect is part of the push. I also expect that millions are not going to be willing to be priced out of "freedom on the road".

I don't disagree that planes are safer than cars, though I want to point out that the trend for less deaths is across transit options. I think modern accident investigation, safety culture, technology, and design all play roles.


By Dennis Bratland - Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=66179446

Apparently it never warned them, so I would conclude they were never in a stall or approaching stall. One of the post here mentioned 3 AOA sensors. The best possible fix. On all the other systems the three vote. The fix that needs to be done will involve the Air Data System (I'm not sure if the 737 has only one Air Data System. It probably has two). It won't be a quick fix. I'll bet the quick solution will be to deactivate the code with a patch and even that would have to be vigorously tested. Right now some engineers are sweating bullets. Engineers always want more testing, while management live by schedule. It would have been way cheaper to eat the penalty money and test more.

The stick shaker was apparently going off the whole time because of the faulty AOA sensor. Gave the pilot's conflicting indications about the state of the aircraft.

Now that appears to be a harder problem to fix. No wonder the crew was baffled. The shaker activated with the nose down. This new "safety" option, is embedded pretty deep. In a future emergency like this, they could "pull" the air data CB's. They lose lots of info, but still have the standby horizon and airspeed indicators. They would be getting their data directly from the pitot system. Plus the turn and slip indicator. One problem though, if they don't train by pulling CB,s they might lose valuable time searching for them. From my memory, most are on the panels behind the crew and one of them would have to turn around. Modifying the code in one of the boxes, will take time. Got to locate the problem, and I would guess the engineers already have a possible fix, BUT, there will be strenuous offline testing, then testing on a simulator, then somehow rigging up a newly updated box for flight testing to get it certified.

that a lot of industrial production equipment depends on sensors. It then becomes a job of monitoring the monitors. Twelve years ago, I worked in a highly technical semiconductor factory, and between movements of "work in progress" from machine to machine, we spent a lot of our time on preparing things that would test the performance of the machines, and thus the sensors that controlled them.

If things got out of whack, we called a process engineer, who sometimes told us to shut the machine down, do a process that works out to be a reboot of sorts, then test again. Another test failure, then another engineer would do repairs, and we'd try again.

All these lane sensors and auto brakes on cars. And the sensor on airplanes that have become numerous over the past 20 years or so.

Oh, wait.

Never mind.

The pilots in question didn't get adequate training about this stability augmentation system. The checklists they're given to handle emergency situations didn't cover these particular failure modes well. The software in that stability augmentation system should be more idiot-proof, and have better behavior, like detecting sensor faults instead of flying the plane into the ground.

Overall, this is a clusterfuck, and something that needs a better fix than "replace the sensor that triggers the bug that kills planefuls of people."

Boeing's working on a software update. That can't come soon enough, but it had better go through better testing this time.

The question is "will it kill people at a rate less than what is already happening today."

For the automakers the answer is apparently a resounding yes because they are pouring MASSIVE amounts of research $ into making this a reality.

And the "get in the car and do nothing" level of technology is not coming anytime soon. It will be phased in over many years with different levels of autonomy.

This is a really good article about it I read last year...

http://discovermagazine.com/2018/nov/baby-can-you-drive-my-car



Home
»
November
»
The Road to Self-driving Cars Is Full of Speed Bumps

FROM THE NOVEMBER 2018 ISSUE
The Road to Self-driving Cars Is Full of Speed Bumps
Automakers are revving up for a very near future of fully autonomous vehicles. But the road ahead is rough.
By Hannah Fry|Thursday, October 25, 2018
RELATED TAGS: TRANSPORTATION, ROBOTS
Self-driving-car
Neil Webb

The sun was barely above the horizon on March 13, 2004, but the Slash X saloon bar, in the middle of the Mojave Desert, was already thronging with people.

The bar is on the outskirts of Barstow, a California town between Los Angeles and Las Vegas. It’s a place popular with cowboys and off-roaders, but on that spring day it had drawn the attention of another kind of crowd. A makeshift stadium that had been built was packed with engineers, excited spectators and foolhardy petrol heads who all shared a similar dream: to be the first people on Earth to witness a driverless car win a race.

The race had been organized by the U.S. Defense Advanced Research Projects Agency, or DARPA (nicknamed the Pentagon’s “mad science” division). The agency had been interested in unmanned vehicles for a while, and with good reason: Roadside bombs and targeted attacks on military vehicles were a major cause of death on the battlefield. Earlier that year, DARPA had announced its intention to make one-third of U.S. ground military forces vehicles autonomous by 2015.

Up to that point, progress had been slow and expensive. DARPA had spent around half a billion dollars over two decades funding research at universities and companies in the hope of achieving its ambition. But then came an ingenious idea: Why not create a competition? The agency would invite anyone in the country to design their own driverless car and race them against each other on a long-distance track, with a prize of $1 million for the winner. It would be a quick and cheap way to give DARPA a head start in pursuing its goal.