4-Dimensional Global Map

One of the challenges for successful Autonomous Vehicles is knowing where to go, following paths to get to there, and avoiding running into things along the way.

An article in the latest issue of Wired magazine on Mapping Mogadishu reminded me of ideas I had about 5 years ago on mapping. In Mogadishu the chaos has destroyed all the records of: people, property, taxes, businesses, everything. The hope is that by building a comprehensive map encompassing all these data, life will return to normal more quickly in Mogadishu.

My ideas were triggered by figuring out how a Personal Transport Vehicle would help my father get around in Charlestown. For example, getting to his table in the dining room required moving through a maze of people, tables, chairs, wheelchairs, and other Personal Transport Vehicles.

The advent of ubiquitous GPS and online maps has dramatically changed how we drive, but we still have to use a lot of concentration and knowledge to translate the line the GPS blithely draws on the screen into a continuous stream of steering, acceleration and braking controls to keep from joining the 40,000 annual dead or the 1,000,000 annual injured in car collisions.

Innovations like traffic prediction and alternate routing are still in their infancy. Recently, my GPS predicted a 12 minute delay on the highway I was just about to turn onto, so I accepted its recommendation to turn onto a side street instead: that took me down a 25 mph road, past 2 schools, which were just letting out; it took more than 12 minutes, and as I turned onto the highway, supposedly just at the end of the congestion, the highway was clear as far back as I could see.

But GPS and online maps aren’t any help at all for my father’s Personal Transport Vehicle getting to his table for dinner. The GPS doesn’t work very well indoors and its accuracy is at best a few feet, which isn’t very helpful when trying to get within 1 inch of the table. Further, the online maps don’t say anything about the layout of the hallways and doors inside Charlestown, much less the placement of tables and chairs, and even less where people are at any instant in time.

You may be thinking: that’s true, but the sensors on driverless cars will take care of that. Good thought, but it doesn’t work for several reasons: line-of-sight, accuracy, 3-dimensionality, and time or the 4^th-dimension.

Line-of-sight: when looking at a mass of people, you can’t see what is hidden behind other people or other obstacles. (Radar isn’t very good at localizing people, and I don’t think blasting everyone with continuous radar will be very popular anyway.) Thus the path you need to follow is probably not even visible from where you are standing, unless you are very tall and the crowd isn’t very big. J

Accuracy: you probably don’t want vehicles coming within about 6” of each other or of people, or even farther at more than walking speed. And my father wants to get within an inch of the table. That resolution requires good optics, with a steady mount and other technical characteristics, such as a long baseline for stereoscopic vision. Not impossible, but challenging.

3-dimensions: as opposed to the 2-dimensions needed to follow a road, where you can assume clearance above and to the sides, we have to consider heights. The Personal Transport Vehicle needs to decide how close to the table to get so my father can eat. Will the arms of the Personal Transport Vehicle fit under the table? Are my father’s hands in the way? We’re talking less than an inch, and in all 3-dimensions, not just a path on the floor.

Time, the 4^th-dimension: how fast does the situation change? Other Personal Transport Vehicles are moving at up to 10 mph (15 feet/second). People are moving half that fast. The path to my father’s table can change dramatically as other people jostle for position to get to their tables – have you ever seen the start of the dinner hour at a senior facility? J

Time is even more complex because we have other events to consider than just local motion: what time does dinner open and close? Does my father have a reservation for a specific time? Is he meeting people for dinner? Are they on time, or early or late? Based on past experience how long will it take to get to the dining room from his apartment? How many other people have reservations at the same time? Is there a party or other event going on in the dining room to make it more crowded, or to draw people to some other site? If he is late will they be out of his favorite dish?

As we expand this situation to include other vehicles and Convoys at much higher speeds, and other types of events and locations, you can see that this is a major challenge.

So what we need is a detailed 3-dimensional model of the world plus continuous updates to include moving and moveable objects. We also need to keep data on the past to predict things like traffic levels, on-time data for scheduled transport, and your favorite trips so you don’t have to keep entering them.

What about the future? We want to know about schedules and reservations and menus for restaurants; when businesses are having sales on particular items; and a myriad other things.

So you can see why I'm proposing a detailed 4-Dimensional map, as essential for helping my father, and all the rest of us too.

Let’s start with ideas for the location challenge, and later I’ll expand to the other challenges.