Learning Domain Randomization Distributions for Training Robust Locomotion Policies

Mozifian, Melissa; Higuera, Juan Camilo Gamboa; Meger, David; Dudek, Gregory

doi:10.48550/arxiv.1906.00410

Cited by 3 publications

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Parameter inference may also be integrated in a closed-loop system (Chebotar et al, 2019;Mehta, Diaz, Golemo, Pal, & Paull, 2019;Mozifian, Higuera, Meger, & Dudek, 2019;Ramos et al, 2019), where the currently estimated posterior over simulation parameters guides domain randomization and policy learning. Such approaches iteratively reduce the sim2real gap.…”

Section: Parameter Inference For Simulatorsmentioning

confidence: 99%

DiSECt: A Differentiable Simulator for Parameter Inference and Control in Robotic Cutting

Heiden¹,

Macklin²,

Narang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Robotic cutting of soft materials is critical for applications such as food processing, household automation, and surgical manipulation. As in other areas of robotics, simulators can facilitate controller verification, policy learning, and dataset generation. Moreover, differentiable simulators can enable gradient-based optimization, which is invaluable for calibrating simulation parameters and optimizing controllers. In this work, we present DiSECt: the first differentiable simulator for cutting soft materials. The simulator augments the finite element method (FEM) with a continuous contact model based on signed distance fields (SDF), as well as a continuous damage model that inserts springs on opposite sides of the cutting plane and allows them to weaken until zero stiffness, enabling crack formation. Through various experiments, we evaluate the performance of the simulator. We first show that the simulator can be calibrated to match resultant forces and deformation fields from a state-of-the-art commercial solver and real-world cutting datasets, with generality across cutting velocities and object instances. We then show that Bayesian inference can be performed efficiently by leveraging the differentiability of the simulator, estimating posteriors over hundreds of parameters in a fraction of the time of derivative-free methods. Next, we illustrate that control parameters in the simulation can be optimized to minimize cutting forces via lateral slicing motions. Finally, we conduct experiments on a real robot arm equipped with a slicing knife to infer simulation parameters from force measurements. By optimizing the slicing motion of the knife, we show on fruit cutting scenarios that the average knife force can be reduced by more than 40% compared to a vertical cutting motion. We publish code and additional materials on our project website at https://diff-cutting-sim.github.io.

show abstract

Section: Parameter Inference For Simulatorsmentioning

confidence: 99%

DiSECt: A Differentiable Simulator for Parameter Inference and Control in Robotic Cutting

Heiden¹,

Macklin²,

Narang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Parameter inference may also be integrated in a closed-loop system [11,54,62,73], where the currently estimated posterior over simulation parameters guides domain randomization and policy learning. Such approaches iteratively reduce the sim2real gap.…”

Section: Parameter Inference For Simulatorsmentioning

confidence: 99%

DiSECt: A Differentiable Simulation Engine for Autonomous Robotic Cutting

Heiden¹,

Macklin²,

Narang³

et al. 2021

Preprint

View full text Add to dashboard Cite

Robotic cutting of soft materials is critical for applications such as food processing, household automation, and surgical manipulation. As in other areas of robotics, simulators can facilitate controller verification, policy learning, and dataset generation. Moreover, differentiable simulators can enable gradient-based optimization, which is invaluable for calibrating simulation parameters and optimizing controllers. In this work, we present DiSECt: the first differentiable simulator for cutting soft materials. The simulator augments the finite element method (FEM) with a continuous contact model based on signed distance fields (SDF), as well as a continuous damage model that inserts springs on opposite sides of the cutting plane and allows them to weaken until zero stiffness, enabling crack formation. Through various experiments, we evaluate the performance of the simulator. We first show that the simulator can be calibrated to match resultant forces and deformation fields from a state-of-the-art commercial solver and real-world cutting datasets, with generality across cutting velocities and object instances. We then show that Bayesian inference can be performed efficiently by leveraging the differentiability of the simulator, estimating posteriors over hundreds of parameters in a fraction of the time of derivativefree methods. Finally, we illustrate that control parameters in the simulation can be optimized to minimize cutting forces via lateral slicing motions. We publish videos and additional results on our project website at https://diff-cutting-sim.github.io.

show abstract

“…Leveraging Bayesian Optimization, Muratore et al [131] update the randomized parameter distribution to maximize policy performance in the target domain. Mozifian et al [132] takes a different perspective by updating the parameter distribution to balance conservativeness and robustness. However, the real-world applicability of this method is uncertain, as it doesn't consider real-world performance explicitly.…”

Section: Learning Robust Policy Through Domain Randomizationmentioning

confidence: 99%

Reinforcement learning for robot assembly

Vuong

View full text Add to dashboard Cite

how to do good research and given me insightful advice and constructive criticism that are distilled into this thesis. I am grateful for his continuous support during my PhD, not only in the lab but also in life.I am grateful to Thao Doan, Huy Nguyen, and Quang-Cuong Pham for introducing me to scientific research. This journey wouldn't have started without their support. I've been fortunate enough to have amazing friends and colleagues in CRI group:

show abstract

Learning Domain Randomization Distributions for Training Robust Locomotion Policies

Cited by 3 publications

References 9 publications

DiSECt: A Differentiable Simulator for Parameter Inference and Control in Robotic Cutting

DiSECt: A Differentiable Simulator for Parameter Inference and Control in Robotic Cutting

DiSECt: A Differentiable Simulation Engine for Autonomous Robotic Cutting

Reinforcement learning for robot assembly

Contact Info

Product

Resources

About