Part 22 - Back to the future

The metaphorical wheel has turned full circle once again.

The road map for autonomous vehicles has long been seen as a progressive path with driver assist features improving the driving experience and making it easier for the driver. This automation takes away many of the decisions, smoothes out the ride makes it all just that much safer.

This of course is a huge improvement and does make driving easier and safer but unfortunately with people being what they are there are unintended consequences emerging.

Once the novelty of the tech wears off, human nature kicks in and the little voices in our head say .. if the car can do it, has done it for months, therefore I trust it to keep doing it.

This mindset has proved fatal in one instance with a Tesla and has had many others blaming the vehicles because they, as the driver, weren't paying attention.

To be fair, this is a pretty valid argument. If the manufacturer has a feature it should work consistently, shouldn't it?

This is the catch 22. It is beta software in the Tesla and it does work but you do need to keep an eye on it all the time, just in case.

However drivers don't pay attention all the time even when driving without any driver assist features as the road toll and panel beaters will tell you. So with very sophisticated driver assist features people become even more complacent.

This is the core problem that Google has been talking about for some time and they have the philosophy that an automated vehicle should be fully automated, no pedals, no steering wheel and no way of the driver taking control.

Other manufacturers are starting to think along the same lines with Ford announcing they will have a fully autonomous vehicle for mass transport use on the road by 2021 and are looking to implement it through ride sharing services.
This is less than 5 years away.

This is a nice argument but does put a lot more pressure on the manufacturers because, once this is released there is no turning back and the technology has to be spot on right from the start.

Statistics will prove that the number of collisions, deaths and injuries will reduce significantly but even so, once someone dies in a fully autonomous vehicle all bets are off and the public trust will be severely eroded. At this point there will be a severe swing by the fringe to roll it back and there will be a lot of debate.

Many people are saying that this technology should be abandoned.
(Sorry ignore the previous line, I've just been reading US political stories).

So this is the conundrum....
Do we continue with driver assist and its limitations or do we go the big bang theory and get rid of all controls?

Interestingly enough some companies appear to be going this route now but keeping it simple and using it in controlled environments.

Companies such as Navya, Easymile  and Ford have autonomous buses for use in private environments usually business parks or universities where the environment is more sedate, low traffic densities and the speed of the vehicles can be restricted. This gives a real world but much safer testing environment. However when this extends to road conditions it faces the same challenges as other autonomous vehicles although speed limitations conceivably could stay lower than normal traffic rates.

Whichever way it goes with automation, there will be a considerable overall reduction in road carnage either way and we have to hope that the big picture takes precedence over the hysteria.

Part 21 - The state of the future

Singapore is a modern clean city state and I have been fortunate enough to visit there on several occasions over the years.

I first went there in 2002 and found it to be very quiet, sedate, very very clean, neat and very easy to get around with low traffic and a fast efficient MRT rail system.

On my last visit two years ago I found it to be far more vibrant and it has grown considerably since my last visit.

Given that it is a city state with limited land resources of 719 sq km (278 sq miles) and a population of more that five and a half million people it has its own set of unique problems .

Not the least of these is traffic.

Given that a basic Corolla on the road with all government fees and charges will cost over $US100,000 in the first year, it is staggering that the government has had to take steps to reduce the number of vehicles it will allow on the road for the next year because of traffic volume and the long transit times around this small country.

As I said I have been there a few times and have noticed how much longer it now takes to get somewhere by taxi compared to when I first visited.

So where is this leading?

Singapore is the ideal place to be the first to implement a fully autonomous taxi service.

Why you ask? Good thing you did really, so now I have a reason to continue.

Singapore is just off the equator so it doesn't get snow or ice. It has a good road infrastructure, relatively flat land and a government with good control and able to implement new policies when they need to.

To tackle the traffic problem the government has committed to have autonomous taxis on the road and has already started testing with Delhi automotive.

With taxi fares expected to drop by a third with the removal of the driver and dramatically increased traffic flow it seems an ideal solution for this small but progressive state.

Their ambitious time frame starts with testing this year (2016) and concluding in 2020 with a complete fleet of fully autonomous taxis on the road replacing the existing manual fleet.

Given the known phenomenon of taxis disappearing off the streets when it rains (everyday in Singapore) this should also help considerably in getting people moving.

Given that this is a very ambitious time frame, it remains to be seen if this happens as quickly as they hope but given the commitment and need, Singapore could well lead the world in implementing this technology.

On another related front, Singapore's SMRT Corporation is to provide automated transport pods for use in  airports, campuses, resorts and industrial parks.These can carry up to 24 people at speeds up to 40 km per hour.

Many countries are talking a lot about autonomous vehicles,
Asian countries are doing a lot more than talking about it, possibly because of greater population pressures and growing economies creating a more urgent need.

The west has had private ownership of vehicles for a very long time but many Asian countries have only recently reached the level of wealth where individuals can afford to buy cars and they are doing so in ever increasing numbers.

This has created a boom which in turn has created huge congestion and pollution issues which puts enormous pressure on governments to keep this growth sustainable.

Automated vehicles may be the key to having efficient transport for all and to keep pollution down by reducing the number of vehicles on the road.

So what seems to be tight time frames to us in the west may actually be achievable in the east....

After all, they have a very big incentive to make it work.

Part 20 - Australia, that big island down south, you know, the one with the Kangaroos.

It looks like we are finally getting there.
Parts of Australia are getting into the swing and realising that this is happening here as well.

Automated vehicles are coming down under.

The Australian Road Research Board Principal Behavioural Scientist Paul Roberts has stated that automotive technology is a lot more advanced than people realise and goes on to say that they will be here a lot more suddenly than people realise. The full story is here:
The Australian arm of Carnegie Mellon University in South Australia is planing to build a research centre in conjunction with General Motors Holden to make Australia an export hub for this technology so it's good news for all of us in the antipodes.

We might be a long way away from the rest of the world but we do have some very clever people, if you discount politicians.

I have been saying for some time that many people have under estimated the speed at which this will happen and when it does, it will appear to happen virtually overnight to those that aren't expecting it!

Coming back to a common theme of the past few weeks, we need to look a little deeper at the question of what makes a vehicle fully autonomous.

The normal answer to that is when you can hop in, tell it where to go and sit back and relax, read a book or watch a movie or even become involved in more intimate activities as some other pundits have suggested.

That is fully autonomous.

That then begs the next question. Is the tech behind it safe enough?

At this point there are no standards.

There are a number of ways of sensing what is around the vehicle, be it cameras, radar, lidar and cloud based GPS systems or a combination of all four.

Obviously the more systems you have the less chance of something being missed and when weather conditions are bad the use of all of these can make it as safe as houses to keep driving.

Therein lies the rub. If you base it on one system then you will never have more than driver assist, as somethings either won't be detected or can be incorrectly identified and human intervention is required to analyse the situation. If the driver isn't alert then the system fails and the "driver" pays the consequence.

This was seriously highlighted recently with the death of a Tesla driver, presumably letting the car take full control. With only cameras, the car didn't see a light coloured truck crossing in front of it against a bright sky.

To be clear, autopilot is a great driver assist feature but is a single technology enough when lives are at stake to allow this as the basis for a fully autonomous system?

To use a bad analogy, would you jump out of an aeroplane with only a single parachute with no backup emergency 'shute?

So we come back to the same question, who decides how many systems need to be used and how many of each?

Cost, obviously is a factor but if you buy in bulk then the cost comes down. Tesla model 3 so fr has advance orders in excess of 300,000, so if you have 2 lidars on each that's 600,000 units straight off the bat. Build in the cost for a full production run over the life of the vehicle model and we are talking millions of units.

Given that new products have to cover the development cost in the first few years then they are expensive. However once the development costs have been recovered the manufacturing cost is far lower and economy of scale kicks in and the wholesale price becomes much cheaper.

Why not then have systems in the vehicle from the start. It will add to the base price of the vehicle obviously but think about who will be the early adopters.

These will be the big fleet buyers for taxis services who are lining up to corner the market as soon as the tech is approved for use, and these players have very deep pockets and are prepared to pay a premium to have the safest systems on the road right from the beginning.

This covers the recovery costs and economy of scale starts to kick in from these clients and the rich private buyers and by the time it trickles down to the rest of us plebs, the costs will be far lower. That is if we even need to buy a car at that point but that's another discussion altogether.

Hopefully there will emerge an international body that sets industry standards and we can get the best possible results, we're talking of saving millions of lives a year after all and every single life counts, especially if it's my family.





 

Part 19 - A day in the life of.

I woke at my normal time and went through the normal cleansing ritual (the three SSS's) and turned on the news as I started breakfast.

Average sort of news day, war in the middle east, former superstar dies at 90, politician charged with corruption, nothing out of the ordinary except for the one story at the end that suprised me and I would think, many others.   

In Sydney, there was a guy on his way to work and was involved in a collision when a tyre blew and his car swerved into another. Fortunately there were only minor injuries and the road was cleared by Sydney automated Crash Clear in a few minutes. Traffic control central issued an automated alert and all other vehicles changed lanes and continued with diversions to allow the emergency vehicles through in priority mode.

Most people were only delayed for a minute or two at most, but the unusual point was that this was the first collision that involved human injury in the entire city in over a month.

  

With the passing of the new autonomous vehicle law and its enforcement last year, it meant that manual driving within the metropolitan area was now illegal. Since then the collision rate, and particularly the death and injury rate had dropped to historic lows.

I finished eating and voice called Valet. "Leaving for work 5 minutes"

I cleaned up, picked up my gear and walked out the front as the car pulled in to my drive.

It changed to my favourite colour and spoke as the door opened. "Good morning George, heading for the office as usual"? 

I grunted an affirmative and engaged the in car display to get an early start on the day.

Having achieved several emails, two video calls and a lot of reading I had clocked up 30 minutes from my work quota before I got there.

I left the car at the kerb as it reconfigured itself for the next user.

I crossed the nearly deserted road to get coffee. 

The coffee shop was a retro kind of place with old posters, one of which was a street scene taken from across the road looking back at this store from about 50 years ago. The contrast was striking. There were masses of cars waiting at the long since removed traffic lights and people waiting for them to change so they could cross. To the side was the multi story car park which is now an open area with gardens and fountains where I quite often spend some time at lunch.

As I get my coffee I hear an ambulance approaching, siren blaring to warn the pedestrians, the light poles start to flash automatically in a pleasing calidescope of colour as they receive the alert to warn the deaf people, and the not so deaf immersed in their music. The vehicle passes swiftly, unimpeded and the sound fades into the distance.

I walk back to my office for the rest of the daily grind.

At 3pm I have Valet call a car to take me to home and I get another 30 minutes work done on the way.

I find a parcel waiting at the door that I ordered online this morning and the auto delivery got it here before I did, better than standing in a queue as I used to do.

Tonight I will have a car pick me up and several friends on the way and we can have a nice evening at a restaurant without having to worry about a "skipper" staying sober to drive home or worrying about finding parking in the city centre.

There was a brawl in the street tonight and the police were there within a minute. No traffic patrols, breathalisers and high speed pursuits for these boys in blue. 

They now get to focus more on real policing and the response time shows, as it does for ambulances and emergency rooms with car crashes out of the scene.

We can all feel safer on the street now on multiple levels.

Valet gets me a car and I can catch a few minutes of my favourite streaming show on the way home after checking and responding to a few private emails.

I reflected as I drifted off that night about how much more time I have now, useful time that I can enjoy with far lower stress levels than I used to have driving or on public transport.

At last Technology has caught up with people and become transparent, where things like Valet have become pretty much a silent servant.

Now technology works for us, we don't have to work to make it work, and it works very, very well.

Part 18 - Today is the beginning of the rest of your life

I know I spoke of this in the last blog but it is a significant event and needs to be looked at in more depth.

Last week the first person died in a vehicle that was driving itself.

Lets start by saying that any death for whatever reason is lamentable.

Suprisingly, the fact that a vehicle was being driven autonomously and caused a death has not started a backlash as the pundits had forecast.

Why is that?

The simple answer is that it wasn't an autonomous car in the first place. It was a great car with a lot of driver assist features that are available within the vehicle to be used in test mode with the drivers full attention, ready to take control.

The driver apparently was using it as an autonomous vehicle, having gained confidence in its features over a period of time - which may have generated a lethal overconfidence when a situation occurred that was outside its capabilities and the driver failed to notice.

The official enquiry will bring out the facts so any speculation at this point is really futile.

Many cars on the road currently have similar features with different systems and sensors all looking at the world differently.

One fact that is indisputable is that this technology is not ready to allow it to be used exclusively as an autonomous vehicle and Tesla is quick to point out that fact.

Unfortunately, if a progressive development model is rolled out over time with more and more driver assist features which still rely on the driver to be awake and aware then that creates a whole new problem.

Imagine you're driving on a long trip and the the road is long, straight and extremely boring, a common scenario in a place like Western Australia where I live, where it can be many hundred of miles - between trees, and even further between towns with the only thing to break the monotony is the odd kangaroo or emu.

But what if the kangaroo or emu decides it wants to occupy the same space on the road as you do at the same time? (not the brightest animals in the world, think "shiny thing, lets check it out").
The semi autonomous vehicle says Oh Oh, I don't know how to handle this and throws its virtual hands in the air and screams ITS YOURS!

What happens?

The driver could take a very long time, in high speed travel terms to go from semi comatose to alert, and by then it could be too late for the driver and most certainly for the emu.

So a case can be made for going directly to fully autonomous vehicles but that in itself creates a whole new set of problems.

It is becoming increasingly obvious that fully autonomous vehicles will be on sale and on the road in 2020 and many others will be offering level two or three vehicles which leads to the problem just mentioned.

That is less than 4 years from now and, as per a lot of countries, we are not ready!

So what does that mean for users?

A very good question as there are, as yet, no industry standards, no federal and very limited state regulation in Australia, and very little anywhere else.

So that means if standards are not set for the use of these features are they legal to use?

Within Australia a driver has to be in control at all times under current law, so what are the implications for using a vehicle in autonomous mode? Worse still what will be the consequences if a vehicle is sold as fully autonomous and someone is killed when the law has not been changed.

All of the above means that the vehicles on sale in many places will have to have controls built in to prevent the full use of autonomous features on the road and to put the onus back on to the driver until such times as the law changes, Mercedes are doing this now.

So that begs the question of what needs to be done to get ready for the future today?

First and foremost we need a legal definition of what defines an autonomous vehicle and that needs to be based on standards that can be measured and quantified such as:

1. Standards for the minimum number of sensors required and their location.
2. Standards for the types of sensors required.
3. Standards to define the minimum sensor range of the sensors fitted.
4. Testing scenarios via NATA and/or  ANCAP to determine the efficiency and reliability of the systems.
5. Law changes to allow drivers to be hands off, with the above criteria met.
6. Insurance companies to allow claims against collision when in autonomous mode.

There are undoubtedly many more but these are the first that spring to mind. 

And remember, we have less that 4 years to do it all before the tech hits and starts being used seriously.

On the same theme, Volvo has reaffirmed that it will take "full responsibility for the operation and efficacy of its autopilot technologies" for their autonomous cars when they hit the road.

This will remove the onus from the driver and put it on Volvo provided the vehicle is used in accordance with the manufacturer and the currents law.

Now that's a level of confidence that you really want to see when you're sitting in a ton and a half of metal hurtling down the road with a computer in control and it also gives the emus a fighting chance.

BMW has also set 2021 as the launch date for their fully autonomous vehicle as well with a new partnership with chip maker Intel to supply the processing power and Mobileye technology for the software.

The future is here, nearly, sort of, maybe.

Part 17 - Ethics and automation, rebel without a cause

Another highlight this week in the world of automated vehicles is, yet again, the question of ethics.

Computers don't know what an ethic is, and even if they did it would still be irrelevant.

Think tanks love to get involved in moral dilemmas. What if? .. it doesn't matter what, can a computer handle it in a way acceptable to people?

Really, it is irrelevant for the simple fact that people react differently to computers and much, much slower.
 If you are presented with a moral dilemma when driving, the human will usually hit the brakes as an instinctive reaction and go into brain freeze as they try to sort out what's happening and by the time they formulate a reaction, it's all over one way or another.

A computer will see the danger long before a human would, start to brake, tell other vehicles about the problem and then react to the circumstance. Ethics don't come into it at all.

The outcome, regardless of the action taken will in the vast majority of situations, always be far better from an autonomous vehicle than by a human as the detection and response time is far faster. So yes, automated vehicles in extreme circumstances may still result in death or injury , but the estimated 90% reduction in traffic incidents far outweighs any moral dilemma.

Ethics, nice to talk about but totally irrelevant to the new world of automated vehicles.

Having said that, we come back to the real world and  new dilemma that we will face increasingly as time goes on.

That is the fact that a person has died in a car under computer control.

Was this an autonomous vehicle, no. It is a Tesla which has beta software running on driver assist technology. Unfortunately it appears that people are relying more and more on it as fully autonomous system which it isn't. In this situation all the facts are not yet available as a full enquiry will be conducted to establish what really happened. Was the driver attentive and ready to take back control? What sensors were in use in that particular vehicle and what were the exact circumstances of the collision?
All of this will be discussed ad nausium in days to come but it does highlight a whole raft of issues to be sorted before fully automated vehicles hit the road.

For instance Volvo wants to have fully autonomous vehicles on the road in New South Wales by 2021. (NSW is the most densely populated state in Australia with Sydney as the capital).

Most, if not all states in Australia have a rule that states a licensed driver must be in control of a vehicle with a least one hand on the steering wheel at all times.

Until that law is changed a car can't drive itself without a "human" driver in control.
Change to this law is essential but it's really only the tip of the iceberg.

Firstly what defines an autonomous vehicle?

• What sensors does it need to be considered safe? Does it have radar and Lidar in addition to a camera?
• What is the range of the sensors?
• What tests have they been put through to satisfy the authorities enough to allow them on the road?

These are a few of the simpler questions that need to be answered before laws can be changed to permit these vehicle on the road in fully autonomous mode.
Australia has some of the toughest vehicle standards in the world and this is rigorously enforced and if vehicles don't comply they are not permitted on the road in Australia.

As I have said in earlier blogs, the governments need to start addressing these issues now before the technology use becomes more widespread in defiance of the laws.

If standards are not applied then this unfortunate death may be the first of many resulting from the use of technology not really designed for the way it's being used and a lot of the work done in improving road safety standards will be bypassed for the sake of expediency.

If that happens there may be many more deaths even with the technology as the technology is just what it says on the label, driver assist.

Interesting concept in the brave new world of electric vehicles.

Imagine a world with all of theses batteries on wheels running around. What if you can share the power? Charge during the day with cheap power and sell it back to the grid at night?

An interesting article here that postulates that you could substantially offset a large part of the car cost using this method.

Link this with the concept of power sharing on the road between vehicles and the whole concept changes.

OK, since I started looking at this it has become a fascinating insight into the possible future.

Electric cars can easily become a power source in their own right.  Power charging could come from the road itself, from parking bays or stand alone charging systems that a car can park at, then move from autonomously when charged.

There are limitless possibilities but using it as a storage bank has definite appeal.
Tesla could incorporate their cars into the "Tesla Power Wall".

All very interesting possibilities.

May you live in interesting times.

Part 16 - A fork in the road

A bit of light reading has prompted me to think of the possible "big picture" scenario in the evolution of autonomous vehicles.

It strikes me that there are really a couple of basic ways of doing this.
I hadn't even really considered this until today but reading between the lines there appears to be a fundamental difference in the way that China and the west are approaching things.

The common school of thought is that the manufacturers build cars that basically do all the things drivers do, (hopefully not the bad things  that tend to kill people) and will effectively be a plug in replacement for the driver. Add to that some communications where cars can talk to each other to know about traffic density, road works and other impediments and the car decides which way to go.
The vehicle programs the route and away it goes.

For the sake of argument and my less than perfect interpretation of how I think China could be heading, lets turn that on its head....

China is already installing devices in manually driven vehicles to allow them to communicate between themselves and  infrastructure as a pilot program and to spread across all of China.

Once that's in place, imagine a possible scenario where a state sanctioned vehicle is built. The state has implemented an integrated, centralised transport infrastructure system including a control centre.
The drivers gets into the car and punches in the destination. The car connects to the centralised control centre that knows the position of every vehicle on the road, the location of every road work, every emergency vehicle and its destination and traffic densities at every intersection.

Based on this information it can calculate the optimal route and send it back to the vehicle which then takes off in autonomous mode. Any changes to traffic conditions can generate a new route, long before the vehicle reaches a congestion point.

Now for the good and the bad.

The good.
With this in place before the vehicles actually hit the road means that the real life integration of manually driven vehicles and autonomous vehicles can be made much simpler. Think adaptive traffic lights from a centralised system to even out traffic flow, diversion from collision scenes, optimisation of routes for emergency vehicles including advance avoidance by the autonomous vehicles and possible in car warning to manual cars as part of the infrastructure build. By optimising the entire system as an entity, rather than individual units, the system can be coordinated as a whole making it extremely efficient.

Now for the downside.
Big brother knows exactly where you are, always. It can also hijack you if it wants to. If you wanted to be really paranoid and go the full conspiracy route, it could also direct you to a drive off a cliff and no one would be any the wiser. Unfortunately history has proven that governments are capable of these kind of acts in both the East and the West.

Anyone who is driving manually won't be able to get away with anything. Run a red light, you're done, because the system knows where your car is and what the light status is. This could be a good thing or a bad thing depending on how far they enforce the rules. It could also supply a lever as automation becomes more common to shift drivers away from manual control.

 The middle ground. 
If a system like this was to be contemplated, at least in the west, it would have to be heavily wrapped in safeguards to make it work without being exploited, hacked or otherwise turned into a weapon of mass destruction.

I can't really see the west evolving anytime soon to a system like this, but in China, given the social and political structure, anything is possible.


New news

At opposite ends of the spectrum both Mini and Rolls Royce are both preparing for the future with not only autonomous but electric cars.

Rolls Royce has built what I think is the best concept car ever, surpassing even the  Australian FJ holden concept.

This has to be the best looking car ever, kind of like Art Deco mating with Futurama.
It will probably never see the light of day as a production vehicle but does show the way the thinking is going - and I really really like it...

Mini, on the other hand is looking at merging two worlds.
Where Rolls Royce is, and always will be, looking at private ownership. Mini, on the other hand, is looking at an alternate world where private ownership is optional.
But customising a shared vehicle to reflect the current drivers personality is a really radical and quite inovative concept. the car colour can be changed by the driver.

Who knows, you may even be able to have an account and the car reconfigures itself on the way to you and you see it as "your car" when it arrives.

More news

New technologies have combined and created a new electric autonomous shuttle bus using 3D printing and it's called Ollie.

Built through crowd sourcing it exemplifies many technologies and processes not even dreamed of ten years ago.

Ollie will be giving autonomous rides at the companies introductory event on the new National Harbour Campus today.16/6/2016.