Physics-Based Selection of Actions That Maximize Motion for Interactive Perception
Physics-Based Selection of Actions That Maximize Motion for Interactive Perception
复制标题
基于物理的动作选择，最大化交互式感知的运动
DOI：
--
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
O. Brock
中科院分区：
文献类型：
--
作者：
Clemens Eppner;Roberto Martín;O. Brock
交互式感知方法利用强有力的相互作用和观察到的信号的变化之间的相关性，从传感器流中提取任务相关的信息。可以从一个单一的交互中获得的信息量取决于两个因素之间的权衡：交互代理的技巧和观察者的可访问传感器模态。当观察者和交互者相同时，与环境的接触会产生更丰富的传感器流，但该流仅包含有意义的交互的相关信息。另一方面，观察一个知识渊博的代理确保有意义的互动，但限制了可用的观察类型（例如，触觉信息接触）。我们提出了一种方法，逐步建立模型的铰接对象，允许生成和选择最佳的探索性操作的交互式感知。我们的迭代方法减少了交互感知的预定义运动的需要：机器人可以自主选择接触涉及的动作，揭示了大部分信息，从而导致改进的模型，使更好的操作。
Interactive perception methods exploit the correlation between forceful interactions and changes in the observed signals to extract task-relevant information from the sensor stream. The amount of information that can be derived from a single interaction depends on the trade-off between two factors: the skillfulness of the interacting agent and the accessible sensor modalities of the observer. When observer and interactor are the same the contact with the environment generates a richer sensor stream, but the stream only contains relevant information if the interaction is meaningful. On the other hand, observing a knowledgeable agent ensures meaningful interactions but limits the available types of observations (e.g. haptic information from contact). We propose a method that incrementally builds up models of articulated objects that allow generating and selecting optimal exploratory manipulations for interactive perception. Our iterative method reduces the need of predefining motion for interactive perception: the robot can select autonomously the contact-involving action that reveals most information, which lead to improved models that enable better manipulations.
DOI：
10.1109/icra.2016.7487714
发表时间：
2016-05
期刊：
2016 IEEE International Conference on Robotics and Automation (ICRA)
影响因子：
--
作者：
R. M. Martin;S. Höfer;O. Brock
通讯作者：
R. M. Martin;S. Höfer;O. Brock
用于环境自由度物理探索的基于熵的策略
DOI：
10.1109/iros.2014.6942623
发表时间：
2014
期刊：
2014 IEEE/RSJ International Conference on Intelligent Robots and Systems
影响因子：
--
作者：
Stefan Otte;Johannes Kulick;Marc Toussaint;Oliver Brock
通讯作者：
Oliver Brock