Self-driving cars: adoption of a key IoT device
With Friday’s lecture focusing on the Internet of Things, including a remarkable talk by the co-founder of Tesla, I have chosen to consider the path of adoption of autonomous driving and associated challenges. After all, with autonomous cars predicted to use over 4000GB of data a day [1], I think they will be the biggest IoT “device” out there in terms of size, data consumption, and market revenue.
Autonomous Driving Levels
Even before the path of adoption, however, what does the term “autonomous vehicle” actually mean? In 2016, the National Highway Traffic Safety Administration (under the US Department of Transportation) recognized the levels of autonomy identified by the Society of Automotive Engineers. The following table outlines the differences between the 6 levels (0-5) based on 4 metrics [2]:
- Execution of Steering and Acceleration/Deceleration
- Monitoring of Driving Environment
- Fallback Performance of Dynamic Driving Task
- System Capability
Taken from [2], SAE International and J3016.
To put this into perspective, Tesla cars are currently in Level 3 [3] where a human driven is expected to respond appropriately to a request to intervene. That said, Elon Musk has been very public about the fact that he wants to target an aggressive schedule for Level 5 fully autonomous driving [4]
Path to mainstream Autonomous Cars
So, when and how will autonomous cars become mainstream? Several crashes have been reported by the self-driving car efforts at Uber [5] and Google [6], several of which involve human drivers making the “first move”, whether or not a valid one, in the incidents. With the companies usually backing their self-driving cars and blaming the human driver involved, the key point here is that the low crash-rates promise of autonomous driving may be hindered by incidents involving human drivers (where there is a factor of uncertainty). I thus believe that the rollout of self-driving cars is as important as the technology behind them. Some thoughts on how to manage this are:
- HOV type self-driving car lanes
- Similar to the HOV lanes in states like CA, you could have dedicated, high-speed, lanes for self-driving cars
- Pros: Most of the time cars are in their own lane only interacting with other self-driving cars, human drivers are familiar with HOV lanes and rules surrounding them; no special infrastructure required (apart from lane widening in certain areas)
- Cons: Adding a single lane doesn’t solve the problems of merging/exiting a freeway, or that of small roads
- Lots of cameras
- Documentation may be the key to safety here. Perhaps self-driving companies should take the lead to install extra cameras (on buses/trucks/sidewalks/lamp-posts) to improve the training data for their cars, and also to serve as evidence during incidents.
- State-specific implementations?
- Another, perhaps less popular with companies, way to rolling out mainstream use of self-driving cars would be have states propose their own laws and guidelines, much like was done for drones. It’s probably safe to say that CA would be one of the states leading the way here. This does beg the question, however, of what happens for cross-state travel?
What about the fun of driving a fast car?
In closing, here are two final thoughts that surfaced during my research for this post:
Will the brand of your car matter?
In a world with mostly self-driving cars, how will the auto industry look? Before the dawn of automobiles, there was likely a whole market around the breed of horse and carriage at various price-points. But how does a self-driving Toyota compare to a self-driving BMW, particularly if car-ownership becomes a thing of the past?
What about the fun of driving?
Some people love the speed and experience of driving. But as human-driven cars become a rarity, will this become a forgotten please? Or perhaps “cars for humans” will become an expensive niche product (much like Swiss watches in the age of $10 Casio digital watches) that only the rich can drive in private race tracks…?
References:
[1] http://mashable.com/2016/08/17/intel-autonomous-car-data/#Qg9.kulNQqqp
[2] https://www.sae.org/misc/pdfs/automated_driving.pdf
[3] https://www.fool.com/investing/2017/07/14/teslas-big-moves-in-self-driving-cars.aspx
[4] https://electrek.co/2017/04/29/elon-musk-tesla-plan-level-5-full-autonomous-driving/
[6] https://www.theguardian.com/technology/2016/sep/26/google-self-driving-car-in-broadside-collision-after-other-car-jumps-red-light-lexus-suv
3 comments on “Self-driving cars: adoption of a key IoT device”
Comments are closed.
Great post Gaurav, I had seen / heard references to the different levels of automation in self driving cars previously but had never actually looked up what distinguishes the level from each other.
Beyond that I think you raise a lot of interesting questions about how to integrate self driving cars into a system where they will clearly be in the minority on roads (at least to start). I had a discussion about this recently with a co worker but it was more around how the AI itself should act. For example, the first self driving cars could be programmed to drive aggressively (though safely) and so be a bigger benefit to those who buy them at the cost of PR. On the other hand if self driving cars are very timid then drivers on the road may begin to take advantage of them (for example, you could always cut off a Tesla because you know the AI will break for you).
http://www.businessinsider.com/aggressive-drivers-bully-self-driving-cars-autonomous-vehicles-study-lse-goodyear-2016-10
Your post approaches this from a higher level. I live the idea of HOV like lanes but am having trouble coming up with an answer to who would pay …
Based on what is already happening in Michigan and California it seems likely that state governments will play a big role here, and differ significantly in their approaches.
http://fortune.com/2016/12/09/michigan-self-driving-cars/
Good post! I feel excited when imagining the prospect of self-driving cars. Self-driving car study is so promising that we can visualize the auto-auto’s future not far from us. However, I’m wondering if a self-driving car induces an accident that causes casualty, whose responsibility is it, the human passengers’ or the car companies’? Will the insurance industry adjust to cover the damage?
Good post! I feel excited when imagining the prospect of self-driving cars. Self-driving car study is so promising that we can visualize the auto-auto’s future not far from us. However, I’m wondering if a self-driving car induces an accident that causes casualty, whose responsibility is it, the human passengers’ or the car companies’? Will the insurance industry adjust to cover the damage? I’m taking the one unit course.