Wednesday, January 25, 2017

Tesla and the future of the autonomous cars

I have been following companies like Tesla and SpaceX for quite some time. Elon Musk has an amazing ability (or sense) to identify which established mature industries are ready for some shake up. It is fascinating to watch how companies like Tesla with no prior experience come and disrupt their corresponding industries. Being "out-innovated" should be one of the biggest fears and yet many large organisations (both in the car manufacturing and space technology industries) didn't get that sense of urgency. They were too slow. This could be due to various reasons - organisational culture that doesn't support innovation, the cozy feeling that you are safe because there are no new players to change the status quo or the sense that the entry barrier is too high. 

What makes Elon Musk's approach unique I think is partially due to his software industry background. The "release early, release often" mantra, that is so common in the modern software development world, allows his teams to iterate at a higher pace (compared to other players), receive valuable feedback and close the feedback loop by releasing newer versions faster than their competitors. It might be OK to release new products/models once a year in the classic auto manufacturing scenarios (with major overhauls every ~4 years) but in the software development world that would be a crime. And we can certainly see the software industry influence in Tesla's approach. The pace of innovation at Tesla is just crazy. They "will never stop innovating".

Earlier this month Elon Musk announced that the new revision for Autopilot HW2 will be rolled out to the first 1000 Tesla cars. All other Teslas that support this update were promised to also receive it but it would run in the shadow mode. This means that it won't be actively controlling the car but it will be collecting stats on the background (what it would have done if it were enabled).



As you can see from the 2nd tweet, the plan was to enable this mode for all cars by the end of this week. With just a minor delay we received this update:



So just after a couple of weeks of testing all Tesla cars with the HW2 package started receiving the update. By running the system in shadow mode on the first test group of cars, Tesla engineers certainly were able to collect valuable real life data that helped them to calibrate sensors. But as you can see from the last tweet, some cars won't be able to complete the camera calibration process and will have to visit a service station for the "adjustment of camera pitch angle". In parallel (according to Musk), engineers are working on a software solution to this issue. Preparing a software patch to mask/compensate for a hardware problem is a very common approach in the IT industry - this is another indicator of what is in Tesla's DNA.

And now this tweet:

As more data arrives, Tesla can keep tweaking and adjusting the capabilities of the new system. This is a spiral. Releasing updates in relatively short 2 to 6 weeks iterations is very similar to the agile approach, that has become a common practice in IT. It delivers increased value to the customers (end users) sooner.

Late last year Elon Musk has announced an ambitious goal for 2017 - a Tesla car (fully autonomously - no touch) should be able to drive from LA to New York, drop the driver (passenger? occupant?) off at Times Square and then automatically drive away to find a parking spot. We live in the future!

As a side note - sometimes it feels like Elon makes a public announcement first and then forces his employees to stick to these commitments. It's a nice way to maintain the high pace.

Another important news this week came from the National Highway Traffic Safety Administration (NHTSA). 

In May last year a 2015 Tesla Model S operated in the Autopilot mode collided with a tractor trailer in Florida. In June 2016 NHTSA opened an investigation PE 16-007 to "examine the design and performance of any automated driving systems in use at the time of the crash." They have finally released a report after completing the investigation of this crash.


 The result of the investigation?

NHTSA’s examination did not identify any defects in the design or performance of the AEB or Autopilot systems of the subject vehicles nor any incidents in which the systems did not perform as designed.
This is great news for Tesla. But, this is not all. This report also highlighted the safety benefits of Tesla’s Autosteer features. Automatic Emergency Braking (AEB) technologies include the following: Forward Collision Warning (FCW), Dynamic Brake Support (DBS), and Crash Imminent Braking (CIB). 

In 2016 20 car manufacturers (99% of the US new car market) made a voluntary commitment to make AEB "standard on virtually all light-duty cars and trucks with a gross vehicle weight of 8,500 lbs. or less no later than September 1, 2022"

IIHS research shows that AEB systems meeting the commitment would reduce rear-end crashes by 40 percent. IIHS estimates that by 2025 – the earliest NHTSA believes it could realistically implement a regulatory requirement for AEB – the commitment will prevent 28,000 crashes and 12,000 injuries.
12 thousand injuries (!!!) with some of them inevitably being fatal. Imagine how many people will make it safe back home to their loved ones - thanks to this amazing technology. A 40% reduction in rear-end collisions alone is a pretty big deal!

Human driver typical reaction time ranges from 0.7 sec to 3 sec with 1.5-2.3 being the average. 0.7 seconds is an eternity in the world of computers. Remember when automatic gearboxes were first introduced? They were slow and clunky. Purists were advocating for the manual gearboxes - better control of a car, faster launch times (quartermile) etc. Technology kept evolving and then suddenly one day we saw models equipped with the auto gearboxes showing faster times. We just have to accept the fact that computers are faster. The same is happening with the Autopilot technologies. 0.7 seconds is plenty of time for a computer system.

Image courtesy of https://www.technologyreview.com/s/534981/car-to-car-communication/


Now combine this with the car-to-car communications and capability gap becomes even wider. Cars will be able to update each other and provide situational awareness. A car driving 2 cars in front of us in the same lane sees an obstacle and initiates emergency braking. A human driver wouldn't be even aware of this emerging situation yet. Compare this with our autonomous car that immediately receives this information and responds by priming the brakes, reducing engine power and preparing to slow down and stop (actively measuring the distance to the car in front of us).

These recent successes and overall progress in the world of self driving cars got me pondering - how will these technologies change the world around us in the near future?

Faster reaction time means that autonomous cars can travel at faster speeds and potentially closer to each other (to increase the flow rates). This will most likely result in dedicated lanes for autonomous cars only (human drivers will not be allowed to drive there). These lanes might also be equipped with additional features to aid navigation and safety.

Since autonomous cars will be safer, the insurance premiums for the human drivers will go up. Do you want to enjoy driving? You will need to pay more for this "luxury". Imagine telling your grand kids stories how you were driving cars yourself - manually - and that those cars were burning toxic flammable liquids in their engines. 

Companies like Uber will introduce autonomous taxis. Unlike humans that need to rest (or work multiple shifts), new taxis will be able to work 24/7. Eliminating the need to pay human driver salary will drive the cost of this service down (compared to standard taxis).

The idea of car ownership will change too. On a typical business day I need a car 1 hour in the morning to get me to work and 1 hour in the evening to bring me back home. It might be tempting to become a passenger in the autonomous car and enjoy reading or browsing Internet while this car takes you to/from work. 

After dropping you off at work:
If it is a taxi it will drive to service the next customer. 
If it is your own car - it will either automatically find parking (to wait till you need it again in the evening) or it will join the fleet of taxis for a few hours while you work to earn some money for you.

Buying a new car will be fun! You will go to your favourite car sales web site, search for a model you like, complete the transaction.... and the newly purchased car will automatically arrive to your doorstep. "Beep, beep, I am here!"

Private car sales might change too. It might be possible to complete the transaction online (most likely via some sort of escrow service for added safety). But there will be no need to find time to meet face to face for inspection. The car will automatically drive itself to the mechanic of your choice. You will receive an inspection report online. After completing the transaction the car will drive to your home address by itself.

Parking will look different. It will most likely go underground to save space (no need to have huge surface parking lots anymore). It can become more dense too - there will be no need to open up doors and autonomous cars can be guided more precisely to park centimeters from each other (while eliminating chips and scratches)

The cars will automatically visit service stations for scheduled service or necessary repairs. They will  also be able to go to the car wash service and return back home. 

Crowd-sourced navigation services like Waze work well. But the true potential of this approach will be unlocked when connected cars begin talking to each other and to some soft of a central traffic planning and navigation centre. By knowing current locations, planned destinations, and current road and weather conditions it will be possible to select the best routes and optimise overall city traffic flows. 

Navigation planning and car-to-car communications will allow special services (police, ambulance, firefighters) to command all self-driving cars to slow down and move aside to free up a lane. It will also be possible to isolate a section of the road or an intersection for roadworks so that autonomous cars will find alternative routes.

We will see more and more electric cars on the roads. And electric cars are silent. Car enthusiasts (rev-heads) love the engine sounds (which might become extinct soon). That meaty, low pitch sound of a V8 engine... Sorry turbo fours and sixes - your sound just can't match it. Remember how popular were mobile phone ringtones? I predict that in the near future electric cars will come "equipped" with a few standard engine sounds (mainly for the enjoyment of the driver). And there will be an option to buy and download additional sound packs containing any imaginable engine sounds for a small fee.

As a security professional I can also see the new risks. For example, what if someone simulates an emergency breaking signal? It could just be a harmless prank - just to watch all autonomous cars coming to a screeching halt. Or it can have more nefarious reasons. The bad guys may trick a car stop or slow down in a given location to make it easier to attack it. There have been similar attacks demonstrated recently when a low tyre pressure signal was simulated/faked leading a driver to pull over and stop to check tyres.

Potentially someone could try to gain an unfair advantage by transmitting wrong data - e.g. to free up a lane or get the priority when crossing an intersection.

But all these issues aside, I am excited about the future. And thanks to the grand vision of people like Elon Musk this future is becoming a new reality.

I would like to hear your feedback. What are your thoughts? Do you have a self driving already or are you planning to buy one soon? How do you think the world will change? Please leave your answers below.




2 comments:

  1. Elon's approach is like this: first, what's important for humanity? Let's address the important issues.

    Second, think from first order of physics. What would be the best solutions.

    Third, find great smart hardworking people, clearly tell them what you want to achieve, let's work hard.

    Fourth, when we hit issues, lets work 24X7 solve the issues, again apply first order of physics.

    To summarize it: "what's most important for humanity; first order of physics."

    ReplyDelete
    Replies
    1. Thank you Anonymous. I like that approach. "What's important for humanity" is another sign of a great leader. He is focussing not on competing against other business but against important global issues - issues relevant for humanity. They are hard to solve. But this makes it even more interesting. And this view doesn't contradict what I said in this post. I believe that Tesla's software industry DNA helps them to be on the winning side. It is refreshingly different compared to other companies in that industry.

      Delete