Robot Cars

Google recently got a licence to operate automated driverless vehicles on public roads. This begins the transition from manually controlled vehicles to automated robots. We will continue to share the road as we move toward automated traffic, but robots will quickly dominate.

Studies of flocks of Starlings show that each bird pays attention to the seven birds around them. The flock votes many times a second on direction of travel. The birds make simple decisions based on limited information, and complex patterns emerge. The birds are able to travel in tight coordinated formations. Schools of fish are the same idea. They travel, find food, and avoid predators in elegant unison, as a team, and none of the individuals make difficult decisions or perform complicated actions.

It’s called an Agent Based Model. Each individual in the system acts independently. They make decisions based on simple instructions and basic information. Complex patterns emerge and they can be the solution to cool puzzles. Vehicles capable of communicating with one another, gathering information about their environment, running simple algorithms every hundredth of a second, and performing the basic actions required of a steering wheel and two peddles, will revolutionize our ideas about personal transportation. Traffic is a cool puzzle, and it’s solvable.

The difference between computer drivers and human ones won’t be small. My drive to work is about 25 minutes. I reach speeds of 110 kph, but the average over the trip is only 50 kph, because I mostly wait at intersections, slow down, then speed up, and jostle along with other traffic on the various congested arteries of the transportation network. Stuck in a system of controls designed to allow people from ages 16 to 96, with wildly different training, skills, and physical abilities, to navigate a powerful machine amongst others, similarly equipped. It’s not ideal. We’re terrible drivers, you and I. Epic gains in efficiency and safety will be possible when people stop driving.

 Traffic V2

Vehicles know where they are, where they are heading, how large they are, and what actions they are capable of. They also know this information about other Vehicles, because the cars communicate with one another, and that information can be verified by Radar, Lidar, video recognition, accelerometers, sensors of various types, and a traffic system designed to help them. They share their intent with the Vehicles around them, and agree on a course of action based on a common set of rules. The group solves a complicated problem, and no one has to work that hard at it.

Vehicles have self knowledge:

• Physical dimensions
• Location (x, y coordinates)
• Projected path of travel (array of x,y @ time)
• Current Direction, Velocity, and Steering
• Maximum acceleration and deceleration (varies with speed and steering)
• Maximum steering angle (varies with speed)

Car A & B communicate. They navigate themselves via a series of Navigational Nodes, like those used by Google’s mapping and direction services. They exchange their intended paths of travel (along with data about their previous travels, which allows cars to propagate updated mapping information, and develop the most efficient routes of travel.) If A & B have paths that do not interfere , no further communication is necessary.

If their paths do interfere they must arrive at an agreement to alter their futures, either by changing direction of travel, or more likely by altering their speeds so as to pass through the collision zone at different times. The decision process need not be complex, just consistent, and the same for both parties. After safety (not running into one another), your goal would be the greatest efficiency for the group, measured as energy expended, or as time delay, compared to the fastest practical time / path through the intersection.

There’s money to be made converting the transportation industry as oil prices rise, private industry is all over this. I’m excited that technology companies like Google are trying to solve traffic, they have all the resources necessary to crush this problem. This is a transition that could occur rapidly.

Industry has had to create a machine that can operate in the most difficult of conditions. They have to operate surrounded by human drivers, who may do anything at any time, in a system with no controls or efficiencies designed for computer controlled vehicles. If automated vehicles can succeed in demonstrating their safety and value under these worst case conditions, the technology can ramp up quickly. As more vehicles convert to being predictable, highly accurate transportation robots, the entire network becomes more efficient, and the benefits of automatically controlled cars will be conspicuous and overwhelming.

Space dorks everywhere are excited about the Mars rover, whatever man. I’m not going to Mars, and neither are you, but I’m going to have a robot car in my lifetime. That’s awesome.