A robot may collect intrinsic and extrinsic properties of targets from the operator, by actively changing its own states, and by actively changing the target states. This thesis studies the robot's exploration of these properties by changing the target states, in order to model objects and scenes. To model the objects, a novel dynamic process is formulated for interactive object segmentation, and a solution based on particle filter and active learning is developed, thus the robot manipulates and learns the object structures incrementally and autonomously. To build abstract object models from the structural object samples, a multilevel part-based object model is developed by applying latent support vector machine to the training of a hierarchical object structure. / To model the scenes, relational semantic mapping method is developed to describe the scenes with both the objects and various object relations. Relational operators are introduced, in order to build relational semantic maps with relational Markov network, where the robot learns object relations actively and incrementally. To find an object in a map, the operator provides the name of target object and the semantic description of the map, and then the robot instantiates a relational Markov network based on the description and learned parameters. After that, it detects the object with the relational Markov network. / This thesis demonstrates the proposed approach on a robot arm, a humanoid robot, and a mobile robot. The results show that the robots learn target information autonomously by manipulating the target models to build sensor-semantics mappings, and use the information to find and manipulate objects accurately. / 一個機器人可以使用多種辦法來獲取物體和場景的語義信息，例如通過和操作對象的交流，通過主動地改變其自身的狀態，或者通過改變目標的狀態。本論文的研究集中于機器人如何通過改變目標模型來學習物體和場景的信息。 / 對於物體的模型，一種新穎的動態過程被提出，以便描述交互式物體分割，並且提出一種基於粒子濾波和主動學習的方案，這樣機器人可以通過操縱和學習物體來自主了解物體結構信息。為了從這些被分割的物體圖片中建立抽象的物體模型，多層次基於物體結構的模型被提出，並且通過將支持向量機應用到分層對象結構的訓練來學習這個模型。 / 對於場景的模型，開係語義地圖被提出，以便描述由物體組成的場景以及其中不同的物體之間的關係。關係還算符被引入到關係馬爾可夫網絡，以便建立基於闕係的語義映射，這樣一個機器人可以自主和主動地學習物體之間的關係。為了地圖找到物體，操作者可以提供目標對象和基於語義的場景描述，然後機器人實例化一個關係語義地圖，通過已經學習到的物體和關係參數。在這之後，機器人通過實例化的關係馬爾科夫網絡來在場景中搜索物體。 / 這種策略被在多種機器人上面被驗證，包括機器手臂，人形機器人，以及移動機器人。結果表明該機器人可以通過操縱目標物體模型來建立傳感器數據和語義之間的影射，並且使用這樣的語義信息來自主和準確地搜尋和操縱對象。 / Li, Kun. / Thesis Ph.D. Chinese University of Hong Kong 2015. / Includes bibliographical references (leaves 151-163). / Abstracts also in Chinese. / Title from PDF title page (viewed on 07, October, 2016). / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only.
| Identifer | oai:union.ndltd.org:cuhk.edu.hk/oai:cuhk-dr:cuhk_1291487 |
| Date | January 2015 |
| Contributors | Li, Kun (author.), Meng, Max (thesis advisor.), Chinese University of Hong Kong Graduate School. Division of Electronic Engineering. (degree granting institution.) |
| Source Sets | The Chinese University of Hong Kong |
| Language | English, Chinese |
| Detected Language | English |
| Type | Text, bibliography, text |
| Format | electronic resource, electronic resource, remote, 1 online resource (v, 4, 163 leaves) : illustrations (some color), computer, online resource |
| Rights | Use of this resource is governed by the terms and conditions of the Creative Commons "Attribution-NonCommercial-NoDerivatives 4.0 International" License (http://creativecommons.org/licenses/by-nc-nd/4.0/) |
Page generated in 0.0193 seconds