AVs Are Going to Need Much Better Navigation Systems

By March 29, 2016Insights

There is a very popular short story by Jorge Luis Borges called On Exactitude in Science.  In this story, the art of cartography attains such perfection that the cartographers build a map the size of the empire, a perfect representation of the province, boundary to boundary. The following generations didn’t find the map useful and let it dissolve and tatter until its remains could only be found in the farthest reaches of the desert. Essentially, AV maps will need to possess the same level of perfect simulation as the map in Borges’ story before they are given rule over the road.

I’m a lifelong Maphead and the adoption of GPS assist while driving was something I got into quite a long time ago. First with several generations of TomTom devices and more recently through maps on smartphones. One thing is clear; even with today’s sophisticated nav systems, a lot needs to be accomplished before we can depend on these devices to keep us on the correct road and get us to where we want to go without the assistance of a driver. What we need to remember is that we humans were the first AVs. When we drive we look at a map, take information from the world around us, process it through our cerebral cortex, and then make decisions on everything from which way to steer through busy streets, to inclement weather adjustments, to lane closures and HVOs, to even avoiding little Timmy running into the street after his ball. We are now asking vehicles to do the same thing, but we can’t just give a car a 2D map and send it on its way. Maps for AVs will have to be hi-def with laser based measuring for every inch of the roadway. Sensor and camera data will need to be relayed to a network of fleeted vehicles (AVs will need to talk to other AVs not only for live updating, but to understand where they are in relation to each other). Several publications have documented the drive of some rally aficionados who took a Tesla S coast to coast in 53 hours. The car handled all of the work, except for high speed freeway curves. (no, they did not adhere to the speed limits). The car was trained to stay in a lane, but, if it did that, the car would have flipped. So the choice on deviating from the lane or slowing down needed to be made by a human. In this case, understanding of the map relative to the car and when to deviate definitely needs to be programmed in. Just this month, a Google car caused its first crash with a bus while trying to navigate around sandbags in the road. The Google driver reported that he thought the bus was going to slow down so he didn’t intervene. Had the bus been autonomous and able to talk to the car, or had other AVs notified this car about the sandbags, would this accident have happened? Would the sandbags have been routinely mapped, like other man-made obstructions and updated in real-time? In the end, maps are a set of directions for the driver, whether human or autonomous. So what we are really talking about here is navigation of which the map is a tool to facilitate.

In addition to the technologies needed for AVs to map busy highways and cities, nav apps can improve the most in the concept of driver vs. traveler functions. Most nav systems have things that work well and features that just plain don’t work. And there are features on some nav systems that others don’t have. For example, Google bought Waze some time ago, yet, they seem to still run as independent systems. What is the integration process and timetable? How hard can this be?

Most nav apps are pretty restrictive: Find location, enter destination, pick mode (walk, car, even bicycle) and hit go.

A short list of improvements needed for all nav systems:

  • Reflect previous routes, frequented stops, and allow the users to suggest a route.
  • Map multiple in-transit modes: what if I take my car to the train and the train to a bus?
  • The ability to do multiple destinations (sequentially) from mobile like you can do with desktop.
  • A nav app that knows the difference between HOV lanes and regular traffic.
  • Mapping off road: Where you can pull over in case of an emergency. Where driveways and parking lot entrances are.
  • A trip saver function (not just destination) and an accessible cache of most used maps when offline. Google has recently added offline maps for some systems, but it is still kludgy and they recently added a search function for the trip, which is to be applauded: once you put in a destination, you can search for a gas station, grocery store, etc. or even a specific place you want to stop on the way without interrupting the end goal.
  • TomTom and a number of car manufacturer apps have a windshield view where it shows you which lanes to be in for the upcoming exit. They’ve had it for years, yet the smartphone apps have not adopted this function. Why is this?

I do wonder if, once AV nav systems evolve to the point they’re going to, that driving maps will potentially get really good right before they vanish. That there will come a time when we don’t even see or need a graphical map when driving. A point where these maps will only exist as binary code without manifestation and the driving map will go the way of Borges’ map. Why would one need a map with an AV?

Maybe the future of AV nav will be like riding shotgun with an old friend who knows the area better than you, so you just sit back and enjoy the ride.