Jul 1, 2021 · The robot identifies plans whose outcomes would be informative about APF, executes those plans, and learns from their successes or failures.

We apply this active learning strategy to the concrete problem of stacking blocks with a real robot, where the blocks are each unique and have non-uniform mass ...

planner which randomly samples action sequences and selects the one with the maximum expected value. III. ACTIVE LEARNING OF ABSTRACT PLAN FEASIBILITY.

An active learning strategy to learn a model for abstract plan feasibility with a information-theoretic approach to improve data sampling efficiency. (IN ...

... Recently, methods combining learning with classical optimization were proposed to overcome these limitations. Their goal is to speed up TAMP through learned ...

APF model to efficiently plan over abstract action sequences. III. ACTIVE LEARNING OF ABSTRACT PLAN FEASIBILITY. Collecting data on real robot platforms is ...

Duration: 4:59
Posted: Jun 25, 2021

Driess et al. [28] propose to learn a neural model for evaluating the hypothesized discrete actions based on visual images. Noseworthy et al.

Jul 1, 2021 · This work presents an active learning approach to efficiently acquire an APF predictor through task-independent, curious exploration on a ...