Open-source benchmarking for learned reaching motion generation in robotics
| Autor: | Jochen J. Steil, Yaron Meirovitch, Seyed Mohammad Khansari-Zadeh, Andre Lemme, Tamar Flash, Aude Billard |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2015 |
| Předmět: |
Technology
Dynamical systems theory business.industry Computer science programming by demonstrations Cognitive Neuroscience Control engineering Robotics standardized comparisons Benchmarking movement primitive dynamical systems human-like motions Human-Computer Interaction Behavioral Neuroscience Open source learning from demonstrations Developmental Neuroscience Artificial Intelligence Artificial intelligence benchmarking Motion generation business reaching motions |
| Zdroj: | Paladyn: Journal of Behavioral Robotics, Vol 6, Iss 1 (2015) |
| ISSN: | 2081-4836 |
| Popis: | This paper introduces a benchmark framework to evaluate the performance of reaching motion generation approaches that learn from demonstrated examples. The system implements ten different performance measures for typical generalization tasks in robotics using open source MATLAB software. Systematic comparisons are based on a default training data set of human motions, which specify the respective ground truth. In technical terms, an evaluated motion generation method needs to compute velocities, given a state provided by the simulation system. This however is agnostic to how this is done by the method or how the methods learns from the provided demonstrations. The framework focuses on robustness, which is tested statistically by sampling from a set of perturbation scenarios. These perturbations interfere with motion generation and challenge its generalization ability. The benchmark thus helps to identify the strengths and weaknesses of competing approaches, while allowing the user the opportunity to configure the weightings between different measures. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|