Selective Visual and Tactile Sensing for Autonomous Systems Perception and Robotic Applications

自主机器人系统的发展代表了包括加拿大在内的世界各地发达国家的一项战略任务，以保护人口，确保适当的医疗保健和个人服务，为制造业提供竞争优势，并维持经济繁荣。然而，最先进的技术仍然面临着其传感和感知能力的限制。在遇到的挑战中，大量数据的获取很快导致饱和的处理能力，这使得自治系统无法充分利用可用的信息来做出明智的决策。尽管如此，最近出现了一种趋势，旨在将视觉，力和触觉传感器的测量结果合并起来，以开发能够执行灵巧操作的系统，试图更接近人类的能力。拟议的研究重点是开发选择性但准确的传感方法，结合机器视觉和触觉探测，以支持先进机器人应用中的自主决策系统。选择性感测被定义为基于先前获取的相同场景的知识，智能地选择下一次从哪里获取信息，目标是最小化理解场景所需的数据总量。因此，选择性传感将信息收集定位在对机器人操作的成功、安全和鲁棒性具有战略重要性的区域。其主要目标是开发综合感知机器人方法，该方法将联合收割机视觉和触觉感知结合起来，以检测、选择性测量、建模和解释由先验未知性质和物理特性的对象组成的复杂环境。然后，对场景的理解将支持移动的机器人或机器人手臂/手的操作，以进一步选择性地探测或与所发现的对象进行交互。更具体地说，正式的选择性3D成像方法将扩展到支持从多个视角采集数据，并集成不同的传感模式，并从彩色和热成像，深度和触觉数据中进行数据融合。该研究还将研究生物启发的图像和信号处理方法，以实现有效的选择性传感，最后，开发原始的自适应顺应控制方案，以支持灵活的机器人在选择性多模态感觉反馈下与静态，移动的，刚性或可变形物体的交互。该研究将为智能传感和自主系统的基本概念和技术创新做出重大贡献，这些系统对关键机器人应用至关重要，例如辅助机器人系统，智能制造以及安全和危险材料操作。

The development of autonomous robotic systems represents a strategic undertaking for developed countries around the world, including in Canada, to protect populations, ensure proper health care and personal services, provide competitive advantage to the manufacturing industry, and sustain economic prosperity. However, the most advanced technologies still face limitations in their sensing and perception capability. Among the challenges encountered, the acquisition of large amounts of data quickly leads to saturated processing capabilities that preempt full usage of the information available for autonomous systems to make educated decisions. Nevertheless, a recent trend has emerged that aims at merging measurements from vision, force and tactile sensors to develop systems capable of performing dexterous manipulation, in an attempt to get closer to human capabilities.****The proposed research focuses on the development of selective but accurate sensing methods combining machine vision and tactile probing to support autonomous decision systems in advanced robotic applications. Selective sensing is defined as intelligently selecting where to next acquire information from, based upon previously acquired knowledge of the same scene, with the goal to minimize the total amount of data required to understand the scene. As a result, selective sensing orients the collection of information over regions that are of strategic importance for the success, safety and robustness of a robotic operation.***The main objective is to develop integrated perceptual robotic approaches that combine visual and tactile perception to detect, selectively measure, model and interpret a complex environment composed of objects of a priori unknown nature and physical properties. The understanding of the scene made available will then support the operation of a mobile robot, or a robot arm/hand, to further selectively probe, or interact with the objects found.***More specifically, formal selective 3D imaging methods will be expanded to support acquisition of data from multiple viewpoints, and to integrate different sensing modalities and perform data fusion from color and thermal imaging, depth, and tactile data. The research will also investigate bio-inspired image and signal processing methods to achieve efficient selective sensing, and finally, original adaptive compliant control schemes with be developed to support dexterous robotic interaction with either static, mobile, rigid or deformable objects under selective multi-modal sensory feedback.***The research will make major contributions on fundamental concepts and technological innovations for intelligent sensing and autonomous systems that are vital to critical robotic applications, such as assistive robotic systems, intelligent manufacturing, as well as security and hazardous material manipulation.**