ACAT
Learning and Execution of Action Categories
| Coordinatore | GEORG-AUGUST-UNIVERSITAET GOETTINGEN STIFTUNG OEFFENTLICHEN RECHTS
Organization address
address: von Sieboldstr. 4 contact info |
| Nazionalità Coordinatore | Germany [DE] |
| Totale costo | 4˙147˙047 € |
| EC contributo | 3˙150˙000 € |
| Programma | FP7-ICT
Specific Programme "Cooperation": Information and communication technologies |
| Code Call | FP7-ICT-2011-9 |
| Funding Scheme | CP |
| Anno di inizio | 2013 |
| Periodo (anno-mese-giorno) | 2013-03-01 - 2016-04-30 |
Partecipanti
| # | ||||
|---|---|---|---|---|
| 1 | GEORG-AUGUST-UNIVERSITAET GOETTINGEN STIFTUNG OEFFENTLICHEN RECHTS | DE | coordinator | 0.00 |
| 2 |
AALBORG UNIVERSITET
Organization address
address: FREDRIK BAJERS VEJ contact info |
DK (AALBORG) | participant | 0.00 |
| 3 |
INSTITUT JOZEF STEFAN
Organization address
address: Jamova contact info |
SI (LJUBLJANA) | participant | 0.00 |
| 4 |
SYDDANSK UNIVERSITET
Organization address
address: CAMPUSVEJ contact info |
DK (ODENSE M) | participant | 0.00 |
| 5 |
UNIVERSITAET BREMEN
Organization address
address: Bibliothekstrasse contact info |
DE (BREMEN) | participant | 0.00 |
| 6 |
VYTAUTO DIDZIOJO UNIVERSITETAS
Organization address
address: K. DONELAICIO G. 58 contact info |
LT (KAUNAS) | participant | 0.00 |
Mappa
Word cloud
Esplora la "nuvola delle parole (Word Cloud) per avere un'idea di massima del progetto.
Obiettivo del progetto (Objective)
ACAT focuses on the problem how artificial systems (robots) can understand and utilize information made for humans. For example, a statement like: 'Put the nut on the bolt and tighten', which one might find in any assembly instruction sheet, is easily understood by us, but not by any robot. Too much background information is missing: Which tool to use? How to hold the tool? How to operate it? And much more. To make this understandable for a machine, ACAT introduces the concept of an 'action category'. Action categories are extracted from large bodies of human sources (text, images) and will fill in the missing background information. An action category, thus, includes the general computer code to perform the action (e.g. by a robot hand). But, as actions are always situation dependent, it also includes all situation specific information (e.g. how to shape the hand, how to generate the correct pose relative to a manipulated object). This is called action-specific background information. Action categories are obtained by combining linguistic analysis with action-simulation and robot self-exploration. They are assembled in an action-specific knowledge base. The ACAT project will develop a system that uses this knowledge base to compile and execute plans as sequences of individual actions. The execution of these plans by a robot serves as a test-bed for the success of the stored action categories. Repeated execution, error analysis, and correction lead to a life-long update process of the action-category knowledge base. The ultimate purpose of ACAT is to equip the robot – on an ongoing basis – with abstract, functional knowledge, normally made for humans, about relations between actions and objects leading to a system which can act meaningfully. For testing, ACAT uses 'instruction sheets' (manuals), made for human workers, and translates these into a robot-executable format. This way the robot will be able to partially take over human tasks without time-consuming programming procedures improving the efficiency of industrial processes. Thus, similar to computer science, where the development of the first compilers had led to a major step forward, the main impact of this project is that ACAT develops a robot-compiler, which translates human understandable information into a robot-executable program.