Level 5 AV's are just around the corner.  Really?!
Images sourced from unsplash.com & freeaussiestock.com

Level 5 AV's are just around the corner. Really?!

I'm going to start my post with an assertion. Level 5 Automated Driving Systems as described in J3016 (SAE's definitive reference on automated driving systems for vehicles) will not be on Australian roads any time soon despite what we read in media releases and online.

To explain how I got there I'll first consider recent media coverage of Audi's new A8 with level 3 automation - this will get some J3016 basics on the table. By the way, Futune.com coverage of the Audi A8 has a neat video, as does the Audi Germany YouTube Channel.

So, starting with the Audi - not willing to 'just trust' the marketing blurb, I reread J3016 cover to cover. Then I returned to the information available on the new Audi A8.

It will accelerate, steer, brake, and start from a dead stop without driver interaction.

Audi state their 'Traffic Jam Pilot' works in motorway style scenarios with a median barrier at a speed below 60 km/h. Indeed, it seems like Audi's Pilot decides where it can do its magic - announcing “Piloted Driving available” (at least in the prototypes demo'ed in China). What Audi claim their Traffic Jam Pilot can do does indeed measure up as level 3 - conditional driving automation. The limitations are what J3016 call the Operational Design Domain (ODD) - the conditions under which Audi's Pilot is designed to work. The accelerating, braking, steering, stopping and starting are the Dynamic Driving Task (DDT).

So … Audi's claim lines up with J3016 level 3 (anyone disagree?!).

[SAE level 5] permits engagement of the Automated Driving System under all driver-manageable on-road conditions

Now let us consider an announcement by Daimler and Bosch. They aim to have level 4 and 5 Automated Driving Systems in 2020 (ish). Interestingly, this announcement describes SAE level 4 and 5 as "fully automated" and "driverless" respectively, in an urban environment. It would need a very generous interpretation of the SAE source to say this description aligns with the intent of J3016.

As I understand J3016, the key difference between levels 4 and 5 is where the Automated Driving System (ADS) can operate - that is, perform the DDT. Level 4 still has a specific ODD. A good example right now would be the easymile EZ10 driverless shuttle. I'm sure the Daimler-Bosch goal is more ambitious - open road, higher speeds - but its still SAE level 4. The EZ10 highlights other point - a level 4 vehicle can be driverless as long as it operates as intended, within its ODD.

SAE level 5 opens the ODD up to "all driver-manageable on-road conditions". Now I 'tip my hat' to the authors of J3016. They acknowledge that there are some conditions that should not be considered driver-manageable - flooded roads is an example. Living in Queensland - a state of Australia that gets affected by tropical cyclones every storm season - this is a regular event that has on occasion taken lives. Encountering something like this, a level 5 vehicle would "achieve a minimal risk condition" – it would pull over!

I moved on from the description of level 5 in the Daimler-Bosch media release and started to consider just what "all driver-manageable on-road conditions" means. From an insurance perspective, in Australia, a driver-manageable (and insurable) condition includes driving on a beach! Rather than go there, how about we consider dirt roads? Unsealed, gravel … whatever you want to call it.

Let me bore you with some figures. Australia has at least 880,000 kms of public roadway (thank you BITRE Yearbook 2016). Extrapolating from Trading Economics figures I concluded about 53% of that roadway is unsealed - some 470,000 kms (better figures anyone?). I acknowledge the traffic on unsealed roads is a lot lower than on sealed roads. Still, most Australians would consider most (all?!) of those roads driver-manageable. Not counting the four wheel drive tracks in the more remote (and amazingly beautiful) parts of Australia, there is still A LOT of kilometres of unsealed roadway that - by definition - the motoring public would expect a level 5 vehicle to handle.

I've driven on just a small sample of the unsealed roadways Australia has to offer. I imagine an ADS that can handle the ruts, bulldust and corrugations (or is it washboarding in your part of the world?) on a good number of them is a long way off. Lay aside the technical challenge for the moment. Is the investment worth it for auto manufacturers given the much lower traffic volume on those roads?

An ADS that can handle (only) sealed roadways in Australia is still at SAE level 4. That's my view, at the moment, based on what I know. 

Allow me another opinion. Audi's approach of explicitly declaring where its ADS will (and won't) work seems eminently sensible. When I'm traveling in my 2021 model Audi Q7 in rural Queensland and its about to leave the bitumen, the car is going to demand I stop taking in the scenery and take control by announcing "Its your turn to drive mate!" - obviously I've gone for the Hugh Jackman voiced audio alert option.

So, back to my opening statement, level 5 … SAE level 5 … that works under all driver-manageable on-road conditions … will not be on Australian roads any time soon.

So, what? SAE level 4 ADS can have a significant positive impact in cities and urban areas ... if their introduction is well managed (the subject of another post?).

My closing thought is simple - and forgive me for playing with William Gibson's words: the future is nearly here, it just won't be evenly distributed.

John Lambert

Owner at John Lambert & Associates Pty Ltd

6 年

Proponents are starting to back away as the numbers of fatal and serious crashes with AV's grows. And that is even though most AV driving is on high standard/ low risk roads. There is a really long way to go to get anything near to Level 5 - if they ever do.

Gerhardt Otto (MBA, MEngPrac)

Project Manager | Digital & Technology, Engineering | Financial awareness | Stock market analysis | semi-retired

6 年

Interesting read Graham.? Just to pick on one technology. Most ITS systems have Lane Keeping Assist technology. It relies strongly on white lane markings because of the "jitter" introduced on the GPS signals. That result in a position error of 2-15m2 rms. Most position (lane keeping) related problems can be solved if military spec GPS accuracy or broad coverage, differential GPS, can be made available. This may solve the unsealed road issues. But the road network still needs to be mapped into the system and that requires continuous maintenance. Will also need secondary inputs from other sensors to enhance the position confidence levels.

Duncan MacKillop

No Surprise - No Accident

7 年

Fear not Graham, the self-driving car will make only a brief appearance on our roads before being abandoned as uworkable, unsafe and undesireable. Even the mighty SAE have absolutely no idea how fiendishly complex the driving task actually is and yet they continue to act as if automated driving is just around the corner. What nobody understands though is that the vast majority of car crashes occur when people continue to do normal things in abnormal circumstances not abnormal things in normal circumstances. The proponents of self driving cars think that the drivers are the problem that need to be taken out of the loop when in fact the only way the transport system functions is because drivers are firmly in the loop. The ability to spot abnormal circumstances that look like normal circumstances is one that has taken millions of years of evolution to achieve and is not likely to be replaced by some dumb computer chip any time soon.

John Lambert

Owner at John Lambert & Associates Pty Ltd

7 年

The troubles go much further. AV proponents use the 92% of injury crashes involve human error - so AV's will be much safer. But they ignore the fact that the 80% of responsible drivers complete 99.998% of their driving trips without a crash where at the minimum there was $1000 - $2000 property damage crash. Part of the remaining 0.002% of trips includes injury crashes. So the 92% human error figure is completely irrelevant. And interestingly I have seen no analysis of the likely breakdown rate of the AV system - that includes actual failures of computer components and sensors and wiring; sensors being disabled by mud and snow; software decisions that are wrong (hence all the AV failure videos) AND how the AV responds to a breakdown? The Google cars when they reported monthly had an appalling crash rate compared to responsible drivers even though they travelled at 25 mph in 35 mph zones and only operated in non-demanding environments! And drivers were sitting at the wheel so they could take over and did take over in very short times far less than 9 seconds. So NO AVs are not safer! Nor will they reduce congestion in the near or medium term; ....

要查看或添加评论,请登录

Graham Taylor的更多文章

社区洞察

其他会员也浏览了