Tracking Fast Trajectories with a Deformable Object using a Learned Model

We propose a method for robotic control of deformable objects using a learned nonlinear dynamics model. After collecting a dataset of trajectories from the real system, we train a recurrent neural network (RNN) to approximate its input-output behavior with a latent state-space model. The RNN internal state is low-dimensional enough to enable realtime nonlinear control methods. We demonstrate a closed-loop control scheme with the RNN model using a standard nonlinear state observer and model-predictive controller. We apply our method to track a highly dynamic trajectory with a point on the deformable object, in real time and on real hardware. Our experiments show that the RNN model captures the true system's frequency response and can be used to track trajectories outside the training distribution. In an ablation study, we find that the full method improves tracking accuracy compared to an open-loop version without the state observer.

Authors

• James A. Preiss
• David Millard 0001
• Tao Yao
• Gaurav S. Sukhatme

Keywords

• Deformable models
• Training
• Tracking loops
• Recurrent neural networks
• Tracking
• Observers
• Hardware
• Learning Models
• Deformable Objects
• Dynamic Model
• Frequency Response
• Recurrent Neural Network
• Internal State
• State-space Model
• State Observer
• Model Predictive Control
• Nonlinear Control
• Robot Control
• System Frequency
• Recurrent Neural Network Model
• Nonlinear Observer
• System Frequency Response
• Classification Model
• Optimization Problem
• Resting-state
• Finite Element Method
• Long Short-term Memory
• Nonlinear Model Predictive Control
• Extended Kalman Filter
• End-effector
• Soft Robots
• Model-based Reinforcement Learning
• Object Of Interest
• Actuator Limits
• Model Predictive Control Problem
• Linear Velocity
• Complex Exponential