Robust Robotic Fleet Management for Warehouse Operations

科学背景：对食品和杂货的点击配送电子商务服务的需求有所增加。这反过来又增加了在仓库中部署机器人车队的需求。在这个领域有非常商业化的参与者，他们有可行的商业解决方案。在文献中，研究人员要么解决多机器人任务分配问题，而忽略了路径规划，要么解决多机器人路径规划问题，而忽略了任务规划方法。然而，同时解决这些问题可以确保机器人之间有效的协作，共享不同订单的工作量和高效的无碰撞多机器人导航。如果机器人在操作过程中出现故障，或者在拥挤的情况下导致延迟完成分配的任务，那么对性能保证的需求也很高。该项目将探索鲁棒机器人车队管理算法的开发，并在仓库部署时保证性能的鲁棒性。研究方法：学生将研究(i)基于子任务（每个订单中的项目）的任务集群，（ii）集中和分散的混合方法，用于分配拣选和交付集群，以及最好使用市场方法的补货任务，以及（iii）相关的无碰撞，时间和能源效率高的多机器人路径规划（MRPP），使用优先多机器人路径规划和时间窗口预订表。MRPP组件也可以以混合方式解决，以提供鲁棒性保证，并确保在大型环境中操作的大量机器人的可扩展性。培训：从单个机器人的角度将思想和视觉传递到协同机器人的角度理解数学优化问题制定在物理机器人平台上工作开发和模拟环境编程和开发算法理解并将行业合作伙伴的要求纳入研究学术写作和出版向广泛的受众传播研究成果

Scientific background:The demand for click-and-deliver e-commerce services for food and grocery has increased. This, in turn, has increased the demand for deploying robotic-fleets in warehouses. There are very commercial players in the field with a feasible commercial solution. In literature, researchers address either the multi-robot task allocation, ignoring the path planning or the multi-robot path planning, ignoring the task planning approaches. However, simultaneously addressing these can ensure efficient collaborations between the robots in sharing the workload of different orders and efficient collision-free multi-robot navigation. There is also a high demand for performance guarantees if robots fail during their operation or if they end up in congested situations causing delays in fulfilling their assigned tasks. This project will explore the development of robust robotic fleet management algorithms with performance guarantees for robustness while deploying in a warehouse.Research methodology:The student will investigate (i) clustering of tasks based on the subtasks (items in each order), (ii) a hybrid centralised-decentralised approach for allocation of clusters of pick and delivery, and replenishment tasks preferably using market-based approaches and (iii) associated collision-free and time and energy efficient multi-robot path planning (MRPP) using prioritised multi-robot path planning with time-windowed reservation tables. The MRPP component may also be addressed in a hybrid manner to give robustness guarantees and to ensure scalability to a large number of robots operating in large environments.Training:Forwarding the thoughts and vision from a single robot's perspective to those of a fleet of coordinated robots Understanding of mathematical optimisation problem formulation Working on physical robotic platforms Developing and working on simulated environments Programming and developing algorithms Understanding and incorporating industry partner's requirements into the research Academic writing and publishing Dissemination of research contributions to a wide range of audience