Skill parametrization approaches and skill architecture for human-robot interaction

Abstract

There is an ongoing shift in industries from mass production to low-batch-production with highly individualized goods. This increases the programming effort required for the producing machines and robots, which is currently carried out by robot experts. For keeping the production economical, new programming approaches are required, allowing shop-floor workers to instruct robots. One approach is to develop robotic skills, which are pre-programmed software modules that only need to be parametrized by the shop-floor user. In this paper, a new software architecture for robot skills is presented, which aims at robustness and human-robot interaction. In addition, four basic demands on the skill parametrization are described that fasten up the process and increase intuitiveness for the user. We give several examples and implement a screwing skill and a pick & place skill, which are demonstrated in two case studies.

Design, Execution, and Postmortem Analysis of Prolonged Autonomous Robot Operations

“…The flow control software was developed by DLR-RM. It has already been successfully employed in the space-related Spacebot-Camp 2015 [1], and in the industrial project RACELab [4]. We showed that the flow control software scales up to systems with more than a hundred processes.…”

“…To ensure the correct function, the robot levels the ground with the instrument (3). Once the instrument is correctly deployed, the rover waits for an external seismic impulse, which the instrument records (4). The rover repeats the process (6 and 7) at a second location and drives back to the landing unit.…”

ARDEA—An MAV with skills for future planetary missions

“…Many different definitions of the term “skill” exist in the literature. Some focus on the intellectual capability of problem solving (Sussman, 1973), some on the physical abilities of the motor system (Peters, Kober, Mülling, Nguyen‐Tuong, & Kroemer, 2012), and some on all functionalities that may be implemented in a state machine (Steinmetz & Weitschat, 2016).…”

Intuitive Task-Level Programming by Demonstration Through Semantic Skill Recognition

“…Verifying safety and reliability is more straight-forward with skill-based approaches, where a robot expert predefines and generically implements the capabilities of a robot system into specific software modules called skills [10]- [13]. Skills are executable programs with logic for e. g. strategies and failurehandling designed for solving specific tasks (such as pegin-hole), and are crafted by an expert.…”

“…Note that PbD can be used as a specific interaction modality within TLP. That is, most interactions are through a GUI, but occasionally skills will be parameterized by physical interaction with the robot in gravity compensation mode, e. g. to demonstrate the location where drilling should take place [3], [11]- [13].…”