For over a century we have been driving cars, enabling us to roam our surroundings with little effort. Now with the introduction of automated driving, machines will become our chauffeurs. But how about getting us around road construction, or finding a friend in a crowded area? What if you just want to explore and find new places? Will these cars be able to handle such situations, and how can you show your intentions?
Currently there is no middle ground between the car taking the wheel or its driver. This is where Scribble comes in: a haptic interface that lets you draw your way through traffic. You draw a path, and the car will follow, not letting you drive but pilot the car. Scribble lets you help your car when in need, allowing you to wander your surroundings once again.
View a video demonstration here.
Felix Ros - designer
Jacques Terken - mentor
The invention of the automobile has allowed people to explore and wander the world, something that was once seen as a privilege can now almost be viewed as a right; the right to wander our surroundings. Highly automated cars should let us do the same.
So what if a vehicle on autonomous mode encounters a road construction: will it be able to understand which way to pass? Will it understand the road worker's hand gestures to stop or go? Or what if you are picking up a friend at the airport and you don't know their exact location—how will you tell your car to find him? Will it start driving in a pattern to test all different locations, and can it even recognize your friend when it sees him?
Perhaps automation and AI will ultimately let us have actual dialogue with our cars, but for now this unlikely for the foreseeable future. In the meantime, we need a new type of interaction that does not let us fully drive our cars, but lets us pilot them and allow us to do the 'car-like things' we have been used to doing over the last hundred years.
Autonomous vehicles constantly predict a path ahead of them that visualizes where they will be at a certain time in the future. The path is updated and adjusted depending on changes in sensor readings and way-flinging algorithms. This easy representation is often used among engineers who develop these autonomous cars to understand the car's intentions. But what if drivers could manipulate this path; if they could mold it to their will and pilot the car to go places that are yet unknown?
Scribble is a haptic pointing device located at the center of the car that lets you draw and adjust the vehicle's projected path in real time on a display. The display shows the road, lanes, other traffic, and potential hazards. The visual cues on the display can tactually be perceived through Scribble's haptic feedback. The haptic sensation enriches the interaction whilst assisting the driver in drawing a more accurate path. Scribble's force is moderate and can be neglected when doing more complex maneuvers such as drawing off the road or drawing in between lanes.
The path only has to be altered when the traffic situation is too complicated for the automated systems to solve, or in the event the driver wants to force his free will onto the car. The validity of the drawn trajectory path is indicated through color changes. The reason for the path to be invalid could have different reasons but should be easy spot and to resolve with a quick re-draw. When the path is drawn from the road, the car is relying on the driver's judgement that the car can drive this path and will follow with caution, this action might be necessary when navigating around road construction or other hazards.
This drawing based interaction is not driving nor being chauffeured around, it is something in the middle. It leverages the human ability to draw and read visual representations; in this case in the form of a line representing location at a given time. This easy interaction is therefore scalable with the level of automation, an increased level of automation means less interaction is necessary since the system can handle more complex situations.
Rather than driving, Scribble requires the driver to plan his move ahead in time. If successful, this can be quite relaxing for the driver, as they can plan first and then focus on something else until further planning is required. As described by a test user: "Normally you have to think twice, now only once", meaning that he does not have to actually execute his planned action since the car will do this for him, so in the meantime he can "… sit back and relax". Interacting with semi-autonomous systems is like a fluid human-machine dance, it is interfaces like these that could potentially teach us how to become better dancers.
Create a Core77 Account
Already have an account? Sign In
By creating a Core77 account you confirm that you accept the Terms of Use
Please enter your email and we will send an email to reset your password.
Comments
Autopilots are already designed to detect and parse road signs. They can already move to pick up a driver autonomously. A fully autonomous vehicle should reasonably be expected to do just as much, among numerous other tasks. The construction zone and the friend pickup scenarios make no sense.
What a sad way of seeing the world !
The machine is smart, hence the machine is right ! That's crazy! A "stupid machine, say an on-off swith for instance, only deals with a simple command, hence there can be no misunderstanding, it's either on or off. A "smart" machine is expected to read and understand a scenario, and deduct the logical action, which necessarily introduces the possibility of misreading and misinterpreting. As they get more advanced, this issues become less present, but they cannot disappear entirely!
However technically advanced a self driving car may be in the future, why assume that it will know and understand everything? You dismiss the construction site scenario as stupid, but I don't think it is... The car is designed to respond to likely scenarios it is expected to encounter, however, its is almost inevitable that it will at some point be faced with a situation that it cannot identify. The construction site with conflicting information such as road markings and a workman waving you through, might not be understood by the AI ( or maybe more unlikely scenarios ... flock of sheep on the road? truck full of bananas spills on the road? the possibilities are endless!), in which case the car will adopt a default "fail safe" and stop. As a passenger, you may want to indicate to the car that it is indeed safe to proceed, rather than sit they hopeless!
The assumption that as a passenger you have to be passive is ridiculous. Have you never suggested a short cut to a taxi driver or asked them to do a detour to pick a friend up on the way to somewhere? Yes, they could order their own cab, but that would be cold, sterile and boring.
And then there is the subject of humans changing their mind. When i drive (or cycle) somewhere, i will frequently change my mind, stop and go back to look at a shop front or pick up something from a store I go past, or just make a random turn to go and explore somewhere.
If we want a smooth transition to AI, it needs to accommodate our whimsy, liberating us from the boring tasks, but without forcing us to take a route we don't like, even if it is the most logical one.
That said, I agree entirely that the interface itself is rather disappointing, not as intuitive as it should be. And when the images are made in a car with a whopping big steering wheel in front of the driver, I can't help thinking that there may be a simpler manual overdrive ;)
You're thinking roads tomorrow will look the same as they do today.
Is a 5-bar linkage 25% better than a 4-bar linkage?