面向场景监控的全方位移动机器人最短时间覆盖

One major advantage of omnidirectional mobile robots is the high maneuverability. When equipped with directional sensors, such as camera and laser range finders, they are expected to efficiently accomplish scene monitoring tasks. Since the coverage efficiency is vital to find out hidden dangers and reduce possible losses, the minimum-time scene coverage problem with an omnidirectional mobile robot will be investigated in this project. On one hand, for coverage strength modeling, a general homeomorphism-based approach will be proposed, which is applicable to both two-dimensional (2D) and three-dimensional (3D) sensors; Furthermore, inspired by the idea of virtual sensor networks, a multi-objective distributed coverage optimization algorithm will be developed for 2D/3D scenes to derive the sequence of key poses, which considers the tradeoff between the coverage performance and efficiency. On the other hand, for omnidirectional mobile robots with active casters, a hierarchical approach will be adopted to address the single-goal minimum-time motion planning problem in the joint space with multiple constraints, such as velocity/acceleration constraints, singularity avoidance, etc.; In addition, graph theory will be utilized to develop a total-time-optimal multi-goal routing and scheduling algorithm, based on which the robot can be controlled to visit the sequence of key poses to fulfill monitoring tasks for 2D/3D scenes with high quality and efficiency. Finally, the proposed approaches in this project will be validated on the self-developed simulation and experimental platforms for the omnidirectional mobile robot.

全方位移动机器人具有机动性强等特点，当配备视觉/激光等有向传感器时，有望高效地完成场景监控任务。考虑到覆盖效率对于快速发现隐患、减小损失极为重要，本项目将重点研究全方位移动机器人最短时间场景覆盖方法。具体而言，一方面，针对有向传感器，提出一种普适性强的、基于同胚变换的覆盖强度建模方法，可适用于二维/三维传感器；在此基础上，考虑性能和效率等影响，研究基于虚拟传感网络的多目标、分布式二维/三维场景覆盖优化算法，得到期望的关键位姿序列。另一方面，针对主动偏心轮式全方位移动机器人，利用分层递阶的方法，在关节空间中，研究速度/加速度、奇异性等多约束下的单目标最短时间运动规划问题；进一步，基于图论方法，研究整体时间最优的多目标路由调度方法，以控制机器人遍历各关键位姿，完成对整个二维/三维场景的高效、高质量监控任务。最后，基于自主开发的仿真平台与全方位移动机器人实验平台，对本项目提出的理论方法进行验证。

将视觉或激光传感器安装到移动机器人本体上，便赋予了机器人智能感知能力。在此基础上，有望高效地完成场景监控任务，并应用于家庭服务、公共服务、危险场合监控等场合。然而，当前针对有向传感器的覆盖强度建模与优化理论尚不成熟，针对全方位移动机器人的最短时间运动规划与调度问题仍需探索。.针对上述问题，本项目提出了有向传感器通用性覆盖强度建模方法、基于视觉距离的并行分布式覆盖优化算法、机器人拟人行为多层运动规划方法、沿已知路径的最短时间轨迹规划方法，多目标移动机器人路由调度算法。此外，将本项目研究内容进行了拓展，提出了基于新型信息熵的机器人主动环境探索方法、基于先验地图的机器人定位方法、加速度层移动机器人视觉伺服方法。将上述方法应用于室内三维场景监控与自主导航递送，提高了机器人执行任务的效率。.项目按照计划书要求执行，完成了计划书所列研究内容，达到了预期指标。共发表学术论文32篇，包括SCI源刊论文15篇，其中中科院一区论文10篇，IEEE汇刊论文13篇；同时授权发明专利1项，申请发明专利4项，获得计算机软件著作权1项。发布了项目相关内容的2项开源代码。