Task-Parameterized Imitation Learning with Time-Sensitive Constraints (2312.03506v1)

Abstract: Programming a robot manipulator should be as intuitive as possible. To achieve that, the paradigm of teaching motion skills by providing few demonstrations has become widely popular in recent years. Probabilistic versions thereof take into account the uncertainty given by the distribution of the training data. However, precise execution of start-, via-, and end-poses at given times can not always be guaranteed. This limits the technology transfer to industrial application. To address this problem, we propose a novel constrained formulation of the Expectation Maximization algorithm for learning Gaussian Mixture Models (GMM) on Riemannian Manifolds. Our approach applies to probabilistic imitation learning and extends also to the well-established TP-GMM framework with Task-Parameterization. It allows to prescribe end-effector poses at defined execution times, for instance for precise pick & place scenarios. The probabilistic approach is compared with state-of-the-art learning-from-demonstration methods using the KUKA LBR iiwa robot. The reader is encouraged to watch the accompanying video available at https://youtu.be/JMI1YxtN9C0

• The paper introduces a constrained Expectation Maximization algorithm for Gaussian Mixture Models that enforces time-sensitive task execution.
• It extends task-parameterized imitation learning to adapt robot behavior in dynamic industrial environments using Riemannian manifolds.
• Experiments with a KUKA LBR iiwa robot demonstrate improved precision and generalization over standard imitation learning methods.

Introduction to Task-Parameterized Imitation Learning

Imitation Learning (IL) is a popular method for teaching robots new skills by having them mimic human-performed tasks. This method is especially useful for automating repetitive tasks in small and medium-sized enterprises, potentially increasing productivity. The challenge lies in achieving precise execution of tasks at specific times, which is critical in many industrial applications.

The Constrained Expectation Maximization Approach

A novel approach presented in this paper tackles the precision issue by introducing a constrained formulation of the Expectation Maximization (EM) algorithm. This algorithm is adapted for Gaussian Mixture Models (GMM), a probabilistic framework widely used in imitation learning. The new method allows for time-sensitive constraints (TSC), ensuring that tasks like pick-and-place operations are performed with exact timing.

Robust Generalization and Real-World Applications

The proposed method effectively generalizes to variations in the task environment by incorporating Task-Parameterization. This is an extension of the original framework that allows the robot to adapt behavior based on varying scenarios. The paper also ensures the method's compatibility with Riemannian manifolds, significant for dealing with complex data such as orientations in robotics.

Evaluation of Performance

Experimental setups, including simulated and real-world tasks with a KUKA LBR iiwa robot, have been used to evaluate the effectiveness of the new method. Comparisons with standard learning methods demonstrate the proposed approach's advantages in achieving the required precision for industrial applications. Additionally, an accompanying video further illustrates the paper's findings.

Summary

In summation, this paper addresses a critical limitation in current imitation learning methodologies by developing a method that ensures precise execution of tasks at specified times. Through a constrained EM algorithm, the method remains robust and adaptable to various task environments. The research's implications are promising for enhancing the efficiency of robotic automation in industrial settings.