Public infrastructure

It is all around us and if it is working well, we hardly notice it. If the bus stops in front of your house at convenient times and the route also serves your office then it blends into your life. You start to wonder who this infrastructure was really designed for when you have to change trains or buses several times to cross town or roads force you to drive miles out of your way to cross a bridge. We have all run from one side of an airport to the other while trying to catch a connection that’s going to fly halfway back to where we came from and thought; there must be a better way to do this. To get where you are trying to go, you are often forced to make a hodgepodge of different services, that spend your time and money appropriately. I have personally driven to a train station that takes you to an airport where you end up riding a bus to the terminal you fly from. With public infrastructure, there really hasn’t been a better way to do this until recently. AI infrastructure could make a bus route feel like a taxi ride.

Optimized for average demand

Our roads, bus routes, rail lines are all designed to serve the general public which means they are often poorly optimized for individual needs. Bus routes take a lot of human planning and can’t go directly between every address combination in a city. The same goes for airlines and roads. Because infrastructure is expensive, cities put extensive planning into designing and optimizing but that makes them slow to react to public needs. There are also different dimensions to be optimized that are at odds with each other. Adding a new freeway entrance next to your house would make your commute easier but it paves over your favorite park. Public infrastructure ends up becoming a system that is wildly inefficient at actually serving individual needs because it is the product of a general solution. The needs of a city today are likely to be far different than 10 years from now but even things like bus routes are brittle. A route that takes 20 minutes by car and is heavily trafficked during rush hours does not have a public transportation route. Instead, you take 2 buses, a train, stop 90 times and the trip takes you 2 hours 20 minutes.

Portland Oregon public infrastructure Tri Met Map.
Portland Oregon Tri Met Map.

Optimize for specific demand

What if you could re-route your roads for the traffic each day? It would be hard to pour concrete that fast but AI could adjust traffic light timing and change reversible lanes. What if you could re-plan your bus routes every few minutes? Right now, years of planning mean that you are always behind in serving citizen’s needs but AI could dynamically re-route buses and train timings so that the same infrastructure can better serve the community. AI is a great optimization function and it could manage the patterns we already see but react to them instantly instead of going through a committee. Ride-hailing companies are already using AI to offer the point-to-point network we all want from public transportation but because they are optimized for response time, they often ask drivers to move empty across the city for their next pickup creating unnecessary traffic. Using AI to improve public transportation doesn’t involve handing everything over to Skynet. A computer can balance an equation without requiring you to trust machine learning or give up control over how things are implemented. It simply means that you are using a computer to respond instantly and adjust service according to rules that you have given it. Next-generation systems that learn from live data could improve on static AI infrastructure models but it would be hard for most cities to let go of that authority right away.

Example of an AI Neural Network
Example of a Neural Network

How AI infrastructure could optimize aviation

Commercial aviation suffers some of the same issues as bus routes. They are planned months or years in advance and do not adapt to changing conditions. Airlines are also optimized for per flight efficiency, meaning the airline doesn’t care if your route is inefficient so long as all of the seats on each leg of that route were full. They will happily fly you on a full flight from Portland to Denver and then on another full flight from Denver back to Jackson Hole, a route that is 40% further than you actually wanted to travel. Check out the direct private KinectAir flight to Jackson Hole that costs the same as a United Airlines 1st class seat and beats them by more than 4 hours, not including the hours of check-in process. Part of this has to do with unit economics and part with infrastructure. Profit margins on commercial aviation are so thin that every mile flown has to be profitable. Airlines also need gates, personnel, and maintenance at each destination they serve. This means that even though there are profitable direct flights between non-hub cities, the legacy air carriers can’t serve them. Some of these juicy, direct routes are covered by smaller carriers but even they fly fixed routes and can’t capitalize on smaller opportunities.

KinectAir branded Pilatus PC-12
Pilatus PC-12

AI software-defined options 

AI is going to change the airline game. Like with roads and bus routes, it is hard to change the physical limitations but software-defined flight schedules mean that the infrastructure we already have can be used to better serve all of us. Existing carriers would benefit from dynamic routing but KinectAir is taking that three steps further with smaller format aircraft, AI, and machine learning. By using the right-sized aircraft, we can lower our fixed infrastructure needs to the point where a grass strip in the middle of Idaho would do nicely. This opens up millions of new direct flight options in the US and Europe. Now, instead of flying hub and spoke which may double your environmental footprint and only serves the largest cities, KinectAir can fly direct from airports that are part of your local community, run by friends and neighbors.

AI-based dispatch

Our AI engine will run in place of fixed routes to balance supply and demand to better serve travelers economically and environmentally. Initially, AI will make decisions that a human dispatcher could make but at scale.  Software can evaluate millions of schedule permutations per second and offer travelers prices that reflect their travel choices. It will discount flights to encourage customers to make mutually beneficial choices and will instantly help reschedule the fleet to accommodate disruptions. Each time a customer books a flight, it can ensure that our fleet can meet all of our commitments while respecting aircraft maintenance schedules and crew rest requirements. In short, it can react quickly, safely, and predictably to efficiently manage a fleet large enough to provide the level of service we all want to enjoy.

Thinking 10 chess moves ahead

AI takes this to the next level by creating a platform that can also drive sales at scale.  When we let machine learning start to take predictive actions, we can move from being reactive to proactive. Machine learning is a tool that can leverage changing consumer demands that would pose a challenge to traditional programming approaches.  You can think of it as a left brain, right brain combination. AI ensures the safety and reliability of our service, while machine learning provides elegant, problem-solving solutions. This combination is the crux of where our company truly advances the way we think about personal air mobility. Imagine an AI dispatch engine that sends spare parts to Elko, Nevada then advertises flights to see ghost towns so that a crew can stop there to install them while waiting for a snowstorm to clear at their next pickup in Denver. KinectAir is creating AI infrastructure and it is absolutely next level.

  • Post category:Blog