MRI: Acquisition of mobile robots to support indoor navigation and online 3D object detection

如今，移动的智能机器人被用于辅助生活和健康护理领域，因为它们可以陪伴人们，帮助他们，并用于有效地支持各种应用，例如导航。这个名为SemaFORR的项目使用两个配备多个真实世界传感器的机器人，在传统导航假设无法使用时提供认知上合理的解决方案。 机器人配备提供自主导航和真实的时间分类从三维范围数据。该项目突出了认知在导航中的作用，通过推动分类算法的最新技术来增强移动的机器人的图像理解能力。解决多方法搜索，推理学习，表示，在线快速检测和分类，有助于解决具有挑战性的问题的基本问题，该项目涉及提高移动的机器人的图像理解能力。这一目标将通过开发真实的时间分类和对象检测来实现，以帮助计算机和人们解决具有挑战性的问题。在开源标准ROS（机器人操作系统）下运行是构建和测试算法的基础：1。通过整合一种新的建筑方法，使机器人能够学习一个动态的空间模型，图特殊的启示，增强导航与认知合理的决策; 2。通过执行基于阶段的3D分类来增强在线实时对象检测和分类;以及3.通过平衡对象分类的准确性和速度来优化算法。

Mobile intelligent robots are used today in the domain of assisted living and health care, as they can accompany people, assist them, and be used to effectively support various applications, such as navigation. This project called SemaFORR, uses two robots equipped with multiple real-world sensors to provide cognitively plausible solution when traditional navigation assumptions cannot be used. The robots are equipped to provide an autonomous navigation and real time classification from 3D range data. The project highlights the role of cognition in navigation, enhancing the image-understanding capabilities of mobile robots by pushing the state of the art in classification algorithms. Addressing fundamental issues in multi-method search, inference learning, representation, online quickest detection and classification that contribute in solving challenging problems, the project involves increasing the image-understanding capabilities of mobile robots. This goal is to be accomplished by developing real time classification and object detection to help computers and people solve challenging problems. Operating under the open-source standard ROS (Robot Operating System) serves as the base on which to build and test algorithms to: 1. Enhance navigation with cognitive plausible decision making by integrating a novel architectural method that enables the robot to learn a dynamic spatial model that diagrams special affordances; 2. Enhance online real-time object detection and classification by performing a stage-based 3D classification; and 3. Optimize algorithms by balancing the tradeoff between accuracy and speed of object classification.