Abstract
Cars
of the future will know much more about their surrounding environment.
We are looking for ways to use this information to make driving
safer and easier. We are investigating a family of control algorithms
that use this information to provide driving suggestions via force
feedback through the steering wheel. We have developed some prototype
control methods that attempt to take into account how users reflexively
react to guiding forces in time critical situations.
Problem
What
is an effective way to effectively provide lateral guidance using
haptics?
Common problems making this difficult:
Problems
with control methods:
Overshoot
Overshoot
typically happens at sharp corners and causes large reaction
forces that lead to another overshoot and so on.
Oscillation
Naive control methods can create forces that throw a user back
and forth between the walls of a channel.
Problems
with the User's Understanding and Reactions to Forces:
There
are a number of problems involved in making haptic path guidance
usable and informative:
-
The user doesn't understand what the forces mean
- The
user is startled by the forces and fights them reflexively
- The
user understand what the forces mean but wants to ignore them
and do something else
- The
user needs to understand what the forces mean very quickly
Proposed
Solutions:
Path
Guide Control Methods
Spring and Damper
The
user's location is tied to the path by a virtual spring and damper.
This is the easiest path guidance control method to implement
but leads to the overshoot problem in some circumstances.
Look-Ahead
Path Following
This
is a method that has been used for autonomous control and we are
adapting it to haptics [1].
The current velocity of the control point is used to predict where
we will be in the future. The point on the guidance path nearest
to the predicted location is used as a target to steer the user
towards.
Channel
Guide Control Methods
Potential Field
The
user feels a force pushing them away from the sides of the channel
towards the middle. This is an easy method to implement for channel
guidance but can lead to the oscillation problem.
Wall
Offset
The
current velocity of the control point is used to predict the location
of the user sometime in the future. If this predicted position
is outside of the channel, the user is steered towards a location
inside the channel.
'Feelers'
When
a 'feeler' comes in contact with a wall, the user is steered away
from the wall. Similar to the Potential Field method but it only
pushes on you when you are moving towards a side of the channel,
which should eliminate oscillations.
Experimental
Setup
We
will implement the control methods on haptic device called the
PHANToM
by SensAble Technologies.
This gives us more control than a standard force feedback steering
wheel. We will test and optimize the effectiveness of the control
methods using iterative user testing.
References
[1]
Reynolds, Craig. Steering behaviors for Autonomous Characters,
Game Developers Conference 99.
|