Here’s my idea. You take the brain of a Tesla Model 3 and all of the associated autopilot sensors and you install it in one or more intersections. Maybe you add some extra sensors to other locations so it has expanded awareness about traffic which is approaching the intersection.
Then you can do a few cool things.
First, you could have the intersection intelligently direct traffic for optimum flow and minimum wasted energy. It would be aware of not only nearby vehicles, but pedestrians and random obstacles–anything that an autonomous car could be aware of. Think about how many times you’ve stopped at a red light when you didn’t need to, or when if the light had stayed green for just two seconds longer you wouldn’t have had to stop and no other cars would have been seriously inconvenienced. Over the course of a year, a single intersection like this would save thousands of person-hours and many tons of CO2.
You could also have the intersection broadcast the info it has gathered over a V2V network. This would give all nearby autonomous cars additional information which would not only make them safer. The autonomous intersection would have better info than an autonomous car because it has the aditional power of being in control of a finite area with relatively fixed context. That would enable it to be more intelligent about stuff like large non-moving obstacles that might suddenly appear in a roadway. Autonomous cars have a hard time knowing if a big non-moving rectangle is a dangerous road hazard or an innocuous sign.
It would intelligently route traffic to easily allow emergency vehicles through with a minimal disruption of surrounding traffic. It could also prioritize mass transit vehicles like busses, which would reduce the time it takes to get places on a bus and make them a more attractive and feasible option for commuters.
The more of these autonomous intersections you add, the more intelligent and efficient they become. Two adjacent intersections could share knowledge to expand their reach in smoothing out traffic flow. A whole city could optimize traffic on a massive scale. A city-wide grid of these would also be able to provide traffic information to cars and route-planning software that would allow trips to predict with high accuracy when and where traffic would form in the future and route around it.
A city like this would also be able to solve some of the expense issues with lidar on autonomous cars. Lidar scanners are super expensive, which is why Tesla has gone the route of creating an autonomous system that doesn’t use them. But while visual-only autonomy can absolutely be superhuman, lidar provides unquestionably more precise and accurate information about moving objects in a 3D space, especially in dense urban areas with many types of moving vehicles, pedestrians and objects, particularly in intersections. So instead of installing one lidar for every single vehicle in a city, you could have one lidar installed at each intersection, which would make all autonomous cars safer in the city while saving the expense.