Motion Coordination for Heterogeneous Aerial Swarms in Congested Environments

• Applicant: Dr. Wolfgang Hönig
• Subject Area: Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics ; Computer Architecture, Embedded and Massively Parallel Systems
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 448549715

The rapid increase in individual robots' technological sophistication makes it easy to imagine a future in which robotics play a much larger role. Yet, as robots become more prevalent, diverse, and autonomous, it is inevitable that they must interact with each other, whether to avoid collisions or to collaborate on a shared task. Indeed, many aspirational applications like robotic search-and-rescue teams, just-in-time manufacturing, and aerial delivery systems will depend on teams of robots, or swarms, operating together in constrained, potentially dangerous, or even maze-like environments.Although such goals have become more attainable, there is still a wide gap in the knowledge needed to achieve them. Industrial success of multi-robot systems has largely been limited to homogeneous robot teams in well-known environments like planned drone shows and specially equipped warehouses. This is largely a result of the dearth of efficient and safe algorithms to coordinate robots in very close proximity to either each other or to other, potentially dynamic, obstacles - conditions that we refer to as congested environments.This proposal is driven by two key applications in which heterogeneous teams of unmanned aerial vehicles (UAVs) are in particular useful. Our first application is motivated by Industry 4.0, where aerial robots move tools, supplies, or finished parts to other areas in a factory. Our second application focuses on tasks where multiple robots collaborate on individual tasks, such as in construction or clean-up scenarios. In both cases, environments are typically highly congested for cost-, time-, or efficiency-related reasons. Heterogeneous robot teams are preferable, as they tend to be more versatile and can adapt more easily to different work loads.The goal of the proposed work is to investigate novel algorithms for motion coordination for heterogeneous aerial robotic teams that operate persistently and in close proximity to each other. We aim to contribute i) novel centralized, decentralized, and hierarchical algorithms that allow heterogeneous aerial swarms to operate in much closer proximity to each other than existing approaches for single- and multi-robot tasks; ii) validation of those algorithms on physical hardware with the two key applications in mind; and iii) improvements to and maintenance of the leading aerial swarms research testbed. For our algorithm design, we use and improve tools from informed search, imitation learning, and trajectory optimization.

• DFG Programme: Independent Junior Research Groups