Cultural Easing Norms and Tweening Autonomous Mobility

Have you ever tried to cross the street in a different country? What happens when you look the wrong way or fail to catch the driver’s attention? What happens when you do catch the driver’s attention? Does it make a difference in their attention or behavior? Do you ever cross the street in front of a car or truck just trusting they will stop?

As autonomous transport becomes more relevant to getting around, culture and transport norms are going to get mixed up in some very interesting ways. With autonomous vehicles already on the road in California, Florida, and Nevada, how does autonomy deal with non-autonomy and everything in-between? And what about the behavior of animals and people who don’t necessarily anticipate the decision making and trajectory of vehicles? Who’s algorithm do we use, and how to we make less disastrous transitions to autonomy by learning how to react without critical failure?

When cultural norms get translated into machine behaviors, we start to make those social behaviors much more concrete and assumed than we otherwise would. Will vehicles behave like animations, with easing functions directing how the accelerate, stop, and navigate around different kind of objects?

I found some easing equations for animating objects on the screen, created by Robert Penners. There’s a flash-based simulator for testing the different behaviors.  

In matching these functions we can see some that don’t have immediate benefit, like the elastic function that overshoots the target stopping point before coming to rest. That doesn’t make for happy parking.

 

Others, like the easeInOutBack suggests one of those annoying maneuvers similar to when a driver suddenly veers left to gain additional turning distance before making a right hand turn into a driveway.

Accelerating too suddenly into an approach and you are likely to send chills into someone’s spine. Make the approach too jerky and it’s possible the other will be confused.

Functions like easeInOutSine, easeInOutCubic, or easeInOutExpo appear more natural, suggesting the approach, moving more quickly into it, and then slowing the closer they get. Each is a matter of degrees, with some being more subtle and others more direct and immediate.

As the military considers different models of automation, including functional, timeline-based, and information-sharing (pdf 3.8mb), I’m anxious to see how social norms and practices get encoded into the patterns of movement. Later down the line you get even more encapsulated suites of behaviors, such as shy-distance and nanny mode.

I’m curious to see how vehicle bounce would play out. It’s probably worth modeling a variety of these in some multi-agent behavior vehicle simulation if it hasn’t been done already: mixed strategies, same strategies, density and frequency-based strategies.

 

Graphs via http://www.frunder.com/tutorials/flash/rpee/

 

Leave a Reply