About self-driving cars
Listening to the 2nd episode of the OppenheimerFunds' podcast titled “Electronified Cars: This Isn’t Your Parent’s Prius” got me thinking about self-driving cars: is there a way we can engage the technology in its current form? A Self-driving car, (also called automated cars and driverless cars) how many of us would love to own one! And humanity has been trying to make this possible since the 1920s. In 2017, it was reported that about $80 billion has been invested in the development of this technology. The result of our collated effort can be somewhat couched in Peter Thiel’s famous quote: “We wanted flying cars, instead we got 140 characters.”. Little progress has been achieved in the desired direction.
To explain progress so far, consider the table below that illustrates how the Society of Automotive Engineers (SAE) frames vehicular automation.
The world currently somewhere between level 2 and level 3 in terms of vehicular automation. The most advanced of self-driving cars are somewhat autonomous but still require human intervention from time to time. In the best of use cases, some cities have begun testing these cars on their streets. Recently, the Google spin-off Waymo launched their driverless taxi service. We asked for self-driving cars; what we got was a car that requires our attention.
The challenge facing self-driving cars is humans. In a study conducted on traffic accidents between 2014 to 2018, it was discovered that when the self-driving cars were in autonomous mode and driving on their own, 38 incidents occurred while moving. In all but one of those cases, the accidents were caused by humans. The source of this arises from the varied nature of human actions and our ability to interpret them and respond appropriately. Codifying will mean coming up with all possible combinations of non-standard actions a human on the road can make, and getting the machine to respond appropriately.
Appropriately varies and is context based. And there are many unique contexts. City is one. Two of the cities I have driven are Lagos, Nigeria, and Lanzarote, Canary Island. Lagos is a complicated and rapidly changing city, and survival requires being deliberate and very conscious in all your transactions and interaction. This attitude is reflected on the roads and while driving, as people make every attempt to get to their destination by inching their cars forward in the most strategically effective way. Lanzarote, on the other hand, is a holiday destination. People are more relaxed in their approach to driving and are more likely to follow the rules. Driving appropriately mean different things in both cities; driving inappropriately in the wrong city can prove to be dangerous to all road users. And that will be the big challenge of moving to level 3 and onwards to level 5; how do you codify human nuances?
Codifying our unpredictability in an equally unpredictable manner is something I believe we will eventually attain. But the challenge is getting there before public interest begins to wane, and signs are available that it might already have begun. Articles have been written questioning progress made in self-driving technology like this one, calling it an effort in futility. Another one was the Anti-robot vigilantes attack of self-driving cars on the launch of Waymo’s driverless taxi service.
As noted, the problem with driverless cars is the man behind the wheel. Despite our baby step towards achieving the dream of fully autonomous cars, I believe the current level of technology is very much able to address real problems. But the problem is, the driverless car solution is being applied to the wrong problem. Driverless cars are able to operate effectively on high speed, obeying driving regulations and avoiding objects on the road. Maybe we shouldn’t be looking for a complete substitute to a driver; a partial one would do. And we would create the environment where the partial driver substitute can be effective. Speeding has been cited as a leading cause of traffic collision along with other human factors like intoxication, weather conditions and street racing. And highways are the primary location for the most fatal of accidents. And this makes me ask the question….
“what about if only driverless cars are allowed on highways?”
The key word is only. The biggest challenge to operating driverless cars is humans interactions. Let ’s expand on this idea.
What is this idea all about? The idea is to create an environment where driverless cars at the current level 2.5 automation level can operate mainstream. It is known that drivers in urban or developed areas require more cognitive processing to navigate their space than highway driving. Erratic behaviours are more apparent in this region and accidents are more frequent when compared to highways. Despite the frequency, however, they are less fatal. The idea is this could reduce fatal accidents by operating a highway human free. So the car would be require a driver in the cities but must be driverless to operate on the designated highway.
How would this operate? This could be viewed through the vehicular level and the network level. On a vehicular level, a new type of service would be introduced called the vehicle automation industry. This service offered would be to upgrade level 0 cars to level 2 and ensure they are safe enough to operate for a long period without manual intervention. Once completed, the car is issued with a license or a driverless certificate. Because this would be deployed on a large scale, I would assume the cost of retrofitting unit cars with a driverless add-on would be very low.
On the network level, designated highways (or lanes on highways demarcated)for automated driving would be blocked off and access would require presenting driverless certification or license. Once the car enters a designated area, the driver would be required to prove his car has driverless capabilities and enter his destination into the interface available in his vehicle. If the driver doesn’t enter a destination, a default designation (based on collected data of users) would be entered and the car would be off on there wholly controlled by the driverless technology plugged to the network. The driver is allowed to change destination in motion by entering a new destination in the interface ; but the driver has absolutely NO control of the steering wheels, pedars and gears of the car. The car can be retrofitted to be converted into a bed and the windscreens blacked off as the drivers input would not be required. The only control the driver has is via the interface available in the car, which must align with the network before it can be executed. On the back-end, this would mean every car that enters into a designated highway would mean logging into a managed network. The network operators would maintain constant communication with vehicles on their network to monitor the accuracy of their responses to environmental factors and coordinate their movements on the highway. They would ensure all the self-driving vehicles follow road regulations and adjust speed to factors beyond their control like weather, road works or unexpected objects on the road. Once the driver gets to his exit and leaves the highway, his car is disconnected from the network and it becomes a car without autonomous features as these features would be turned off.
And who would benefit from this? On one level you have the companies better able to justify investments and even generate re-investable returns. There is also the new industries that would emerge, from the retrofitters to the ergonomist, seeking to redesign the car for enhanced driverless experience to the network operators.
Back to the question initially asked: is there a way to deploy self-driving technology on a mass scale given our current level of automation? I think the answer is yes.
