Direct Force Feedback using Gaussian Process based Model Predictive Control

Matschek, Janine; Jordanowa, Reni; Findeisen, Rolf

doi:10.1109/ccta41146.2020.9206279

Cited by 7 publications

(4 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…MPC allows tackling a wide variety of force and motion control tasks [12]- [15], [17], [25], [29], spanning from direct and indirect force control to joint hybrid motion/position and force control. We focus on hybrid position and force control, where some positions, as well as specific forces/moments, should be controlled jointly [7], [43].…”

Section: Problem Setupmentioning

confidence: 99%

“…We use Gaussian processes (GPs) [19], [20] to capture and learn the state-force interactions. Force modelling via Gaussian processes is, e.g., considered in [21]- [25]. Learning of friction and grasping forces using GPs is considered in [21], [22], while [23] uses GPs to process sensor data from a tactile array.…”

Section: Introductionmentioning

confidence: 99%

“…Contact forces occurring in human-robot interactions are considered in [24] via GPs. The interaction of a robot with a static environment using GPs to obtain a learned contact force model for MPC is considered in [25].…”

Section: Introductionmentioning

confidence: 99%

“…This mapping is learned using GPs, while ensuring constraint satisfaction on the forces with high probability to ensure safe interaction. Compared to [8]- [17], [25]- [31], we propose to combine first-principle models, e.g., linear spring models, with GP models. The resulting hybrid model is used in a model predictive control scheme for combined motion and force control.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Safe Machine-Learning-supported Model Predictive Force and Motion Control in Robotics

Matschek¹,

Bethge²,

Findeisen³

2023

Preprint

View full text Add to dashboard Cite

Many robotic tasks, such as human-robot interactions or the handling of fragile objects, require tight control and limitation of appearing forces and moments alongside sensible motion control to achieve safe yet highperformance operation. We propose a learning-supported model predictive force and motion control scheme that provides stochastic safety guarantees while adapting to changing situations. Gaussian processes are used to learn the uncertain relations that map the robot's states to the forces and moments. The model predictive controller uses these Gaussian process models to achieve precise motion and force control under stochastic constraint satisfaction. As the uncertainty only occurs in the static model parts -the output equations -a computationally efficient stochastic MPC formulation is used. Analysis of recursive feasibility of the optimal control problem and convergence of the closed loop system for the static uncertainty case are given. Chance constraint formulation and back-offs are constructed based on the variance of the Gaussian process to guarantee safe operation. The approach is illustrated on a lightweight robot in simulations and experiments.

show abstract