GenAI talk for Young at Wageningen University & Research (WUR) March 2024
Robust and Smooth Force Sensor-less Power Assist Control
1. Robust and Smooth Force Sensor-less
Power Assist Control
V. Salvucci S. Oh H. Fujimoto Y. Hori
The University of Tokyo
2. Outline
1 Power Assist Robots without Force Sensor?
2 Proposed Force Sensor-Less Power Assist Control
3 Results
Human Input Based Evaluation
Rigorous Experimental Verification
4 Conclusions
3. Outline
1 Power Assist Robots without Force Sensor?
2 Proposed Force Sensor-Less Power Assist Control
3 Results
Human Input Based Evaluation
Rigorous Experimental Verification
4 Conclusions
4. Power Assist Robots without Force Sensor?
Can we control these robots by estimating the user force with cheap sensors?
5. Why without Force Sensors?
Control key point
Knowledge of user’s applied force
1 Force Sensor
2 Force Sensor-less Control
(encoder based)
Why without force sensors?
Advantages
Low cost
Low weight
Compactness
Faster Response: No delay in force
measurement
Safety: the robot is a sensor
Difficulties
Robustness (ex: friction non-linearities)
Difficult to provide smooth assistance
6. Outline
1 Power Assist Robots without Force Sensor?
2 Proposed Force Sensor-Less Power Assist Control
3 Results
Human Input Based Evaluation
Rigorous Experimental Verification
4 Conclusions
7. Force Sensor-Less Power Assist Control (FSPAC)
Constant gain [Oh 2008]
Based on disturbance observer for:
Disturbance rejection
Force-to-assist estimation
Variable Gain [Salvucci 2010]
VD-TG = Velocity Dependent
Triangular Gain (i.e. variable
impedance)
8. VD-TG = Velocity Dependent Triangular Gain
|va| VD-TG why
|va| < |v1| 0 no assistance stability in unknown environment
|va| |v2| high high assistance at low speed to overcome inertia
|va| ! |v3| decrease low assistance at high and dangerous speed
9. Outline
1 Power Assist Robots without Force Sensor?
2 Proposed Force Sensor-Less Power Assist Control
3 Results
Human Input Based Evaluation
Rigorous Experimental Verification
4 Conclusions
10. Human Input for a “Feeling” Evaluation
No assistance
Low backdrivability
Conventional FSPAC
Not smooth
(oscillatory
movements)
Unstable
Not Robust
Proposed FSPAC
Smooth
Stable
Robust
11. Experimental Apparatus
High Back-drivable System
Actuators: 1 linear motor
Sensors: 1 linear encoder
Motion is transmitted to the door
through the linear motor
! low friction, and high
back-drivable
Low Back-drivable System
Actuators: 1 AC motor
Sensors: 1 rotary encoder
Motion is transmitted to the door
through the ball screw system
! high friction, and low
back-drivable
15. Outline
1 Power Assist Robots without Force Sensor?
2 Proposed Force Sensor-Less Power Assist Control
3 Results
Human Input Based Evaluation
Rigorous Experimental Verification
4 Conclusions
16. FSPAC with Variable Impedance Works
Proposal
A simple-to-implement Variable Impedance Controller for Power Assist
Robots with low cost Position Sensors
Advantages
1 Higher robustness:
Experimentally stable for different humidity and room temperature
conditions (i.e. friction modeling error)
2 Smoother assistance:
Higher assistance to the user at low velocities (=assistance when needed!)
Better input tracking
3 Higher safety:
No oscillatory torque references and velocities that can cause instability
17. Thank you for your kind attention
Hori-Fujimoto Lab
— Eco and Human-friendly Motion Control—
18. References
S. Oh and Y. Hori. Generalized discussion on design of force-sensor-less power assist
control. In Advanced Motion Control, 2008. AMC ’08. 10th IEEE International
Workshop on, pages 492–497, 2008.
V. Salvucci, S. Oh, and Y. Hori. New approach to force Sensor-Less power assist
control for high friction and high inertia systems. In Industrial Electronics (ISIE),
IEEE International Symposium on, pages 3559–3564, 2010.