Patent · US Active

Controlling position of robot by determining goal proposals by using neural networks

US11958529B2 · kind B2 · utility

3Cited by
4References
24Claims
0Family size

Assignee

Inventors

Key dates

Filing dateAug 20, 2020
Grant dateApr 16, 2024
Priority date
Expiry dateDec 29, 2042

Classification

  • Technology area (CPC G)Physics
  • CPC primaryG06N7/01
  • WIPO fieldComputer technology
  • WIPO sectorElectrical engineering

Abstract

A framework for offline learning from a set of diverse and suboptimal demonstrations operates by selectively imitating local sequences from the dataset. At least one embodiment recovers performant policies from large manipulation datasets by decomposing the problem into a goal-conditioned imitation and a high-level goal selection mechanism.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.