Robot Program Parameter Inference via Differentiable Shadow Program Inversion

  title={Robot Program Parameter Inference via Differentiable Shadow Program Inversion},
  author={Bastian Alt and Darko Katic and Rainer J{\"a}kel and Asil Kaan Bozcuoğlu and Michael Beetz},
  journal={2021 IEEE International Conference on Robotics and Automation (ICRA)},
  • Bastian AltDarko Katic M. Beetz
  • Published 26 March 2021
  • Computer Science
  • 2021 IEEE International Conference on Robotics and Automation (ICRA)
Challenging manipulation tasks can be solved effectively by combining individual robot skills, which must be parameterized for the concrete physical environment and task at hand. This is time-consuming and difficult for human programmers, particularly for force-controlled skills. To this end, we present Shadow Program Inversion (SPI), a novel approach to infer optimal skill parameters directly from data. SPI leverages unsupervised learning to train an auxiliary differentiable program… 

Figures from this paper

Heuristic-free Optimization of Force-Controlled Robot Search Strategies in Stochastic Environments

An automatic, data-driven and heuristic-free approach to optimize robot search strategies which adapt to the time-variant characteristics of the underlying probability distributions, while requiring very few real-world measurements is introduced.

Bayesian optimization with unknown constraints in graphical skill models for compliant manipulation tasks using an industrial robot

This article outlines how a hybrid force–velocity controller can be applied to an industrial robot in order to design a graphical skill-formalism and proposes suitable constraint GP models and acquisition functions to obtain new samples to optimize the information gain, while also accounting for the success probability of the task.

Learning to Acquire and Adapt Contact-Rich Manipulation Skills with Motion Primitives

  • Computer Science
  • 2021
Experimental results show that the proposed framework takes less than 1 hour to acquire the insertion skills and as few as 15 minutes to adapt to an unseen insertion task on a physical robot.



Iterative inversion of neural networks and its application to adaptive control

An iterative constrained inversion technique is used to find the control inputs to the plant, and the forward model of the plant is learned, and iterative inversion is performed on line to generate control commands.

Dynamical Movement Primitives: Learning Attractor Models for Motor Behaviors

Dynamical movement primitives is presented, a line of research for modeling attractor behaviors of autonomous nonlinear dynamical systems with the help of statistical learning techniques, and its properties are evaluated in motor control and robotics.

Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

Qualitatively, the proposed RNN Encoder‐Decoder model learns a semantically and syntactically meaningful representation of linguistic phrases.

ACNMP: Skill Transfer and Task Extrapolation through Learning from Demonstration and Reinforcement Learning via Representation Sharing

A novel LfD+RL framework, namely Adaptive Conditional Neural Movement Primitives (ACNMP), that allows efficient policy improvement in novel environments and effective skill transfer between different agents and real-world suitability is shown.

Movement Primitive Learning and Generalization: Using Mixture Density Networks

Representing robot skills as movement primitives (MPs) that can be learned from human demonstration and adapted to new tasks and situations is a promising approach toward intuitive robot pro

Adversarial Skill Networks: Unsupervised Robot Skill Learning from Video

This work proposes a novel approach to learn a task-agnostic skill embedding space from unlabeled multi-view videos by using an adversarial loss, and shows that the learned embedding enables training of continuous control policies to solve novel tasks that require the interpolation of previously seen skills.

PyTorch: An Imperative Style, High-Performance Deep Learning Library

This paper details the principles that drove the implementation of PyTorch and how they are reflected in its architecture, and explains how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance.

Differentiable Gaussian Process Motion Planning

This work proposes a differentiable extension to the GPMP2 algorithm, so that it can be trained end-to-end from data, and performs several experiments that validate the algorithm and illustrate the benefits of the proposed learning-based approach to motion planning.

A Differentiable Programming System to Bridge Machine Learning and Scientific Computing

Zygote is described, a Differentiable Programming system that is able to take gradients of general program structures and supports almost all language constructs and compiles high-performance code without requiring any user intervention or refactoring to stage computations.

Differentiable Algorithm Networks for Composable Robot Learning

This paper introduces the Differentiable Algorithm Network (DAN), a composable architecture for robot learning systems that combines the strengths of model-driven modular system design and data-driven end-to-end learning.