Sailing with the Uncertainty: Dynamic Rendezvous of Under-actuated Vehicles in Semi-structured Geophysical Flows
Abstract
The key of this work is to leverage the stochasticity of under-actuated air vehicles# mobility in semi-structured flow fields to achieve rendezvous. Rendezvous problems in flow fields are usually solved as navigating multiple vehicles in a vector field to realize synchronized arrival at a pre-determined location. Such approaches see limits in real world scenarios due to the complications in sensing and perception, localization capabilities, as well as the navigation difficulties in turbulent flow. Keeping multiple vehicles synchronized while navigating as planned is expensive in both hardware and energy budget. An optimized pre-defined rendezvous location and navigating paths are also hard to select due to the lack of precise knowledge of the environments. Our goal is to pursue acomprehensive solution to simplify the hardware requirements and relax the assumptions of the environmental dynamics, while a team of vehicles can still deliver in-air rendezvous with a high probability.Thisproject aims to develop strategies for air vehicles taking stochastic movements in a continuous space to reach a dynamic rendezvous at a non-specified location with a high probability and infinitely often, and further synthesize time-varying networks via recursively establishing such rendezvous. We see our solution hasunique contribution by dropping the requirements of selecting rendezvous points and vehicles paths, hence rendering the requirements on the mapping of the flow field. Our method leverages both the laminar and turbulent flow structures rather than fighting against the flow, therefore significantly reduce the energy budget. Additionally, the method we proposed avoids hub-like team topology andis resilient to environment uncertainties.Specifically, the objectives of our research are:1)Closing the gap of graph-based models and the real-world environments, such that graph-based methods leading to rendezvous strategies can be applied to real-world problems.2)Develop rendezvous strategies of random-walking vehicles that are: (i) in confined spaces, where vehicles may be physically connected and therefore the distance in between has a hard upper-bound; (ii) taking anomalous random walks, which, in contrast to typical Brownian motion, more frequently used in modeling the motion of objects in turbulent flows; and (iii) in large quantities, where alow probability of rendezvous is compensated by the large number of agents available for rendezvous, and a sequence of multiple rendezvous may happen.3)A 3-D experiment platform composed of individual air-vehicles that can be produced in scales, each can fly for a long time, and can perform both well-controlled and random motions. Our solution will be validated and tested in the Defense of the Republic game (the DTR game), which requires delivering in-air objects through goals, by reaching pairwise rendezvous or a sequence of rendezvous among autonomous blimps.The project contributes to naval research and is of DOD interest in the following aspects:1)Robot swarming techniques: The anticipated outcome includes strategies that will yield emerged rendezvous events among a large number of mobile agents that making decisions upon its own sensing capabilities and very limited knowledge of the neighbors. 2)Coordinating vehicle teams in complicated environments. The methods are suitable for real-world environments where most naval activities take place, and are dark, communication denial (underwater, the Arctic, etc.), or have intrinsic dynamics (the wind, the ocean current,etc.). 3)Resilient systems in risky environments. Ocean environments, especially those parts that are less explored, pose great risks for sensors to deliver tasks in a long term. Long-term operation also requires consideration of environmental adversaries. Our research lead to solutions that can deliver performance successfully under partial failures of the system.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 15, 2023
- Source ID
- N000142312536
Entities
People
- Xi Yu
Organizations
- Office of Naval Research
- United States Navy
- West Virginia University