Interactive Trajectory Adaptation through Force-guided Bayesian Optimization

Abstract

Flexible manufacturing processes demand robots to easily adapt to changes in the environment and interact with humans. In such dynamic scenarios, robotic tasks may be programmed through learning-from-demonstration approaches, where a nominal plan of the task is learned by the robot. However, the learned plan may need to be adapted in order to fulfill additional requirements or overcome unexpected environment changes. When the required adaptation occurs at the end-effector trajectory level, a human operator may want to intuitively show the robot the desired changes by physically interacting with it. In this scenario, the robot needs to understand the human intended changes from noisy haptic data, quickly adapt accordingly and execute the nominal task plan when no further adaptation is needed. This paper addresses the aforementioned challenges by leveraging LfD and Bayesian optimization to endow the robot with data-efficient adaptation capabilities. Our approach exploits the sensed interaction forces to guide the robot adaptation, and speeds up the optimization process by defining local search spaces extracted from the learned task model. We show how our framework quickly adapts the learned spatial-temporal patterns of the task, leading to deformed trajectory distributions that are consistent with the nominal plan and the changes introduced by the human.

Imitation Learning for Variable Speed Contact Motion for Operation up to Control Bandwidth

“…Figs. [14][15][16][17] show the angular responses of θ res s1 and the torque responses of τ res s1 , τ res s2 , and τ res s3 for a height of 5.6 cm. The blue lines represent the responses of the motion-copying system, whereas the orange lines show the responses of the proposed method.…”

Geometry-aware Bayesian Optimization in Robotics using Riemannian Matérn Kernels

“…In addition, when adapting parametric robot policies, some of their parameters such as mean vectors or covariance matrices lie in Riemannian manifolds. This inherent geometry is often overlooked in robotic applications [33,40,41]. In this paper, we demonstrate that incorporating the correct geometric structure of X into the Gaussian process is an important potential avenue for improving performance of Bayesian optimization.…”

Path Learning in Human–Robot Collaboration Tasks Using Iterative Learning Methods

“…In [18], a hidden Markov model (HMM) is combined with Bayesian optimization, which has capability to learn a task with few demonstrations. In [19], Bayesian optimization is combined with PbD to make the robot learn the intended trajectory of human based on the noisy interaction force.…”

“…(3) Different from the machine learning method that requires prior knowledge about the task model [17]- [19], the proposed method is suitable for tasks where the nominal plan of the robot does not conform with a presumed probabilistic model.…”