Transcript IMECE04Accel

Accelerometer Based Handwheel
State Estimation For Force
Feedback in Steer-By-Wire Vehicles
Joshua P. Switkes
Ian A. Coe
J. Christian Gerdes
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
What is Steer-By-Wire?
• Future automotive technology
• No mechanical connection between handwheel
and roadwheels
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Steer-By-Wire
• Steer-By-Wire Advantages
– Reduced packaging constraints
– Increased front impact safety
– New possibilities for vehicle control
• Lanekeeping assistance
• Change handling dynamics
• Challenge: Handwheel and roadwheels
decoupled
– No natural force feedback
• No mechanical transmission of force feedback
• Need to recreate force feedback artificially
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Outline
•
•
•
•
•
Sensing needs
Accelerometer setup concept
Data processing
Simple force feedback implementation
Experimental results
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Sensing Needs
• Steer-By-Wire creates new possibilities for
force feedback
– Complete freedom to tune force feedback
– Remove unwanted high frequency vibration
• Resulting signal has low bandwidth
– Modify mechanical feel
• Damping: need angular velocity of handwheel
• Inertia: need angular acceleration of handwheel
– Tire forces
• Position of handwheel
– Add other force feedback sources
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Sensing Needs cont.
• Steer-by-wire control
– Precise control of roadwheels to track
handwheel requires feedforward control
• Position, velocity and acceleration of
handwheel command
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Automotive
Restrictions/Current Solutions
• Cost of sensing must be extremely low
– High resolution encoder cost prohibitive
• Current Handwheel Sensor Solutions
– Potentiometer
– Low resolution encoder
• Both position based
– Requires double differentiation for
acceleration
• Each derivative increases noise in signal
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Possible Solution: Accelerometers
for Force Feedback
• MEMS Accelerometers and gyros very attractive
– Small (5mm x 5mm)
– Cheap (appr. $1 in quantity)
– Reliable (no moving parts)
• Directly measure velocity or acceleration
– No differentiation needed
– Should result in cleaner signal for same cost
• Still need absolute measurement of position
– Accelerometers have bias which must be removed
– Use cheap low resolution encoder or potentiometer
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Accelerometer Setup
• Two dual-axis 1.7g
units
• Opposed direction
of measurement
allows cancellation
of accelerations
common to two
sensors (gravity or
vehicle
accelerations)
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Data Processing
• Concept:
– Potentiometer removes bias from
accelerometer
– Accelerometer gives clean acceleration
signal and is integrated for velocity and
position
• Implementation
– Kalman filter combines measurement
sources based on knowledge of noise
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Kalman Filtering
• Standard sensor fusion with Kalman filter
• Three states: Handwheel position, velocity,
and accelerometer bias (constant offset)
• Filter dynamics:
Propagate
Kinematics
Group Meeting
Correct for bias
Using Position Sensor
5-Nov-04
Propagate states
From Accelerometer
D
D
L
ynamic
esign
aboratory
Kalman Filter Tuning
• Filter gain chosen using knowledge of system
noise
– Filter is optimal given noise assumptions
• Accelerometer and potentiometer noise
measured in stationary vehicle, with engine
running, handwheel not moving
• Bias variance not measured: tuning parameter
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Experimental Implementation
• 1997 Corvette
converted to steer-bywire
• Belt drive force
feedback on steering
wheel
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Simple Force Feedback
• More complicated feedback possible
– Simple case to evaluate state estimates
• Inertia and Damping recreated on handwheel
– Inertia decreased by conversion to steer-by-wire
setup due to disconnection form massive steering
rack and wheels
– Damping decreased due to disconnection from
power steering, bushings
– Both returned to approximately stock values for
Corvette
• No model involved: direct feedback of state
estimates
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Experimental Results
• Noise sources isolated through testing
– Isolate algorithmic or mounting errors
• Stationary vehicle with steering
– Isolate errors due to road excitation
• Moving vehicle, no steering
– Isolate errors due to vehicle motion
• Roadwheels steering without handwheel
motion
– Fully realistic implementation
• Moving vehicle with driver
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Stationary Vehicle
• Bias variation
due to alignment
error
– Error is easily
estimated and
cancelled
– Industrial
implementation
sufficiently
precise
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Moving Vehicle, Not Steering
• Potholes produce
noise in
acceleration
– Velocity is clean
• Autonomous turns
produce small
error in bias
– Small compared to
normal force
feedback signal
– Easily cancelled
Straight Over Potholes
Group Meeting
5-Nov-04
Autonomous Turns
D
D
L
ynamic
esign
aboratory
Moving Vehicle
• Velocity and
position very
clean
• Acceleration
cleaner than a
twice
differentiated
potentiometer
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory
Conclusions
• Accelerometers can be used to
effectively estimate handwheel states in
a moving vehicle
• Measurement scheme cancels common
mode noise, allowing use in noisy
environment
• Simple low cost way to estimate states
for force feedback
Group Meeting
5-Nov-04
D
D
L
ynamic
esign
aboratory