Michael Ouimet

Coordinated rendezvous of underwater drifters in ocean internal waves

This paper considers a team of spatially distributed drifters that move underwater under the influence of an ocean internal wave. The teams objective is to estimate the physical parameters that determine the internal wave, use the then-known ocean dynamics to rendezvous underwater, and finally return to the surface as a cluster for easy retrieval. From the structure of the internal wave, the oceans flowfield is time-varying and spatially dependent on depth and position along the wave propagation direction. The drifters can control their depth by changing their buoyancy and are otherwise subject to the horizontal flowfield at their given depth. We consider two different drifter dynamical models: a first-order Lagrangian model, useful when the drifters mass is sufficiently small, and a second-order model, where the drag force caused by the water accelerates the drifter. We propose provably correct distributed algorithms that rely on the drifters opportunistically changing their depth so that the ocean flowfield takes them in a desirable direction to perform coordinated motion. The drifters converge to the same depth and position along the wave propagation direction asymptotically. Simulations illustrate our results.

Robust, Distributed Estimation of Internal Wave Parameters via Inter-Drogue Measurements

Internal waves are important to oceanographers because, as they travel, they are capable of displacing mass, such as plankton and small fish. This paper considers a group of drogues estimating the physical parameters that determine the dynamics of an ocean linear internal wave. While underwater, individual drogues do not have access to absolute position information and can only rely on inter-drogue measurements. Building on this data and the knowledge of the drogue dynamics under the flow induced by the internal wave, we propose the Vanishing Distance Derivative Detection Strategy to allow individual drogues to determine the wave parameters. We analyze the correctness and robustness of this strategy under noiseless and noisy measurements, respectively. We also introduce a general methodology, termed pth-Order Parameter Fusion, for combining the parameter estimates obtained at different times and characterize its error. Simulations illustrate our results.

Collective Estimation of Ocean Nonlinear Internal Waves Using Robotic Underwater Drifters

This paper considers a group of drogues whose objective is to estimate the physical parameters that determine the dynamics of ocean nonlinear internal waves. Internal waves are important in oceanography because, as they travel, they are capable of displacing small animals, such as plankton, larva, and fish. These waves are described by models that employ trigonometric functions parameterized by a set of constants such as amplitude, wavenumber, and temporal frequency. While underwater, individual drogues do not have access to absolute position information and only rely on inter-drogue measurements. Building on this data and the study of the drogue dynamics under the flow induced by the internal wave, we design two strategies, referred to as the Vanishing Derivative Method and the Passing Wave Method, that are able to determine the wavenumber and the speed ratio. Either of these strategies can be employed in the Parameter Determination Strategy to determine all the remaining wave parameters. We analyze the correctness of the proposed strategies and discuss their robustness against different sources of error. Simulations illustrate the algorithm performance under noisy measurements as well as the effect of different initial drogue configurations.

Distributed estimation of internal wave parameters via inter-drogue distances

This paper considers a group of drogues estimating the physical parameters of the oceans linear internal waves. While underwater, individual drogues do not have access to position information and instead rely on inter-drogue distance measurements. We introduce the PARAMETER ESTIMATION BY ZERO-DERIVATIVE METHOD for determining the parameters of a horizontally propagating internal ocean wave and show that, for noiseless measurements, our strategy determines the parameters exactly. In the presence of additive Gaussian noise, we characterize precisely the robustness of our strategy and bound the error in the estimated parameters. Finally, we define two strategies for aggregating estimates obtained across multiple time instants. Several simulations illustrate our results.

Heterogeneous Autonomous Mobile Maritime Expeditionary Robots

For the ONR-funded Heterogeneous Adaptive Maritime Mobile Expeditionary Robots (HAMMER) project, we work on cooperative autonomy for a fleet of unmanned vehicles working together in the aerial, water surface, and underwater domains. Each of these systems work well independently, but our goal is to integrate their performance into one system of vehicles that can safely perform cooperative tasks. The challenges we are working on include creating reliable communications links between vehicles in the harsh low bandwidth maritime environment, integrating novel onboard sensors and inter-vehicle communication to create filters to estimate the state of the network, and creating autonomous takeoff-and-landing algorithms between the aerial/underwater vehicles and the surface “mothership” vehicle. The surface vehicle is envisioned to be capable of transporting the aerial and underwater vehicles as well as providing mission-lengthening power. Possible applications of this system include automated deployment and recovery of data-collecting unmanned underwater vehicles and an ad hoc wireless network where the aerial vehicle relays time-sensitive data collected from the surface or underwater vehicle to a human on a ship many miles away. In a separate but related project, we are also determining human-autonomy teaming required for future Naval programs, assessing the state-of-the-art algorithms, and creating open challenge problems to academia to fill gaps based on the Navys need.

Network integrity via coordinated motion of stratospheric vehicles

This paper considers the task of deploying mobile wireless repeaters on aerial vehicles in the upper air of Earth to increase the Internet connectivity of users with little to no network infrastructure. By routing data between a user and a strong connection to the Internet through the vehicle network, this ad hoc infrastructure can provide Internet access to those currently without it. We consider two different types of vehicles, altitude-actuating balloons and fixed-altitude gliders, and examine the task of finding optimal vehicle trajectories under lateral wind dynamics employing a throughput-based performance function. Given the complexity of computing the optimal trajectories of all vehicles simultaneously, we introduce an approximation that allows us to determine the optimal trajectory of a single vehicle. We use this optimal trajectory as a reference for the network and develop a coordination algorithm that makes all vehicles follow it while being spaced out equally in time, providing good overall coverage to users. For gliders, we show that, if cost of control is small enough, these vehicles converge to a stationary formation rather than flow with the wind. Simulations validate the throughput performance of the proposed periodic trajectories.

Robust estimation and aggregation of ocean internal wave parameters using Lagrangian drifters

This paper considers a group of drogues whose objective is to estimate the physical parameters that determine the dynamics of ocean nonlinear internal waves. While underwater, individual drogues do not have access to absolute position information and only rely on inter-drogue measurements. Building on this data and the structure of the drogue dynamics under the flow induced by an internal wave, this paper improves in three different ways upon our previous strategy, termed Parameter determination strategy, which determines all wave parameters. The first is by showing that with sufficiently fast sampling, the extended algorithm determines the wave parameters. The second is by showing its applicability to situations where two internal waves are present simultaneously. With the extended algorithm, multiple estimates are calculated for each parameter. Thus, with the presence of noisy measurements, the third contribution is a method to aggregate parameter estimates to reduce the error. Simulations illustrate the algorithm performance under noisy measurements, the effect that the initial drogue locations has on robustness, and the effectiveness of parameter aggregation.