Holy Grail of Self Driving Vehicles
Tesla and Mercedes and Google and BMW and Apple are all working on the self driving cars of the future. It is a hard problem to solve. In fact, I coined a new term to describe this and called it a "phantom solution."
- July 2016
Technology is a wonderful thing. With each new generation, things seems to get better and smaller and cheaper and smarter. However, when it comes to autonomous driving cars, the solution seems illusive. I came up with a new term to describe this. I call it a "phantom solution."
This is descriptive of the problem. The solution seems to be imaginary. Why is this? The problem is with implementation logistics or how to get universal adoption. Theoretically, if we could switch all cars from manual to autonomous on the same day, the solution would be easy. Whatever software hardware combination we implement, it will work because all vehicles will be in the network. The problem is we can't do that realistically. Any solution we come up with will have to deal with the migration from manual driven cars to autonomous driven cars incrementally. For the most part, there will be a mix of these vehicles on the roads. How will the autonomous cars behave in a world with unpredictable humans behind the wheel?
Getting over the Hump
Anatomy of a Road Trip
To understand the problem better, lets examine a typical road trip. I often take a trip from my house in Yorktown NY to the Westchester Archives in Elmsford NY, a 25 miles trip door to door. I make this trip once a week doing some volunteer work. To get there, I need to backout of my driveway, drive about 3 blocks of local roads, enter the Taconic Parkway going southbound, drive approx. 20 miles, turn onto Route 287 going East, then take exit 3 and drive on local roads for about 8 blocks, turn left at a signal light, go about a mile and turn into parking lot of the Archives. If time permits, I often stop at a local Dunkin Donuts shop and get a coffee along that 8 blocks. On a return trip, I often stop at an Asian groceries store located in downtown White Plains before heading home.
It is a typical trip but it highlights some of the problems that a self driving car would face. The easy part is that 20 miles of highway. It is a 3 lanes highway for the most part with well defined lane markers.
Some problems with the local drives and the parking lots are describe here.
- My driveway does not have lane markers. It is wide enough for two cars. I own 2 cars and sometimes my son come to visit and we have 3 cars in the driveway. In that case, two cars are parked on one side and the 3rd car is parked on the other side. How will a self driving car determine where to park?
- At the Archives, the parking lot is bigger and have stall markers. However, some slots are reserved for employees and some are for visitors, and some are for handicapped. I need to pick the right spot to park. In addition, on some hot sunny days, some spots are shady under a tree. I often will choose those spots even though they may be a little further from the front entrance.
- One of the local road is a heavily used commercial zone with an auto repair shop. Often, there are cars parked temporarily on the road blocking traffic. I often had to maneuver around the parked car by going into the opposite lane. This road only have one lane in each direction.
- The times I stop for coffee, the parking lot of the Dunkin Donuts shop is chaotic. People park sloppily and don't always follow the rules. I use my judgement to determine the best place to park.
- The parking lot in White Plains is a paid lot consisting some outdoors and some indoor spaces. The indoor space is out of range of GPS. It is a congested area where parking spots are not always easy to spot.
In addition, when driving on the parkway, during rush hour traffic, it is quite congested. The speed limit is 55 mph but everyone drive around 65 mph to keep up. There are not too much spaces between each car.
Problem I see with Total Autonomous Cars
In light of the current driving environment, it is hard for me to imagine a car that will take me from door to door without any human assistance. There are just too many unknown situations that can crop up. Such as road construction, downed trees, disabled vehicles, deer crossings...
As with the parking situation, there are instances where I can make the better judgement with regard to parking that is best suited for me. I can't imagine a self driving car making the same decisions without adequate input from the passenger.
The other problem I see is the lack of markers and clearance that often accompany some roads. The sensors and GPS can only resolve distances within limits. There are instances where I have to guide my vehicle around obstacles that only allow inches of clearance.
Being around other human drivers is also a challenge. They will not always abide by the rules of the road. There are times where I must react as a defensive driver to avoid collision or hitting a jay walker or a run away pet. These instances happen on a daily basis and I have little problem dealing with it without a second thought. It just comes with the territory.
All that I mentioned above presents great challenges to a self driving car. There are just too many options and scenarios that require constant re-computation and adjustments. It is the unexpected that offer the greatest challenge and threat to safety. If the system is too strict, it can only lead to congestion and traffic blockage. If it is too lax, it may lead to accidents.
Viable Niche Markets
I do see several areas where autonomous cars can be implemented successfully. For example, the shuttle around airports for rental cars. Right now, they are driven by humans. It is a boring task. The shuttle follow the same path and makes various stops to allow passengers to come on or get off. This makes it a much more predictable scenario. Also, the speed of these shuttles tends to be slower than most cars, around 30 mph.
In some downtown cities, a tour bus makes its round along key landmarks mainly for tourists. This is also an excellent candidate for a self driving bus.
A school bus also falls on the same category where the path is limited and repetitive.
In this hub, I try to define the problem with self driving cars. It is an illusive problem mainly because life is chaotic and no amount of planning can remove them. In order for this system to work, it would have to deal with the bigger problem of coexisting with human drivers. That is my humble opinion.