Search

Understanding the dynamics of bodies of water and their impact on the global environment requires sensing information over the full volume of water. In this article, we develop a gradient-based decentralized controller that dynamically adjusts the depth of a network of underwater sensors to optimize sensing for computing maximally detailed volumetric models. We prove that the controller converges to a local minimum and show how the controller can be extended to work with hybrid robot and sensor network systems. We implement the controller on an underwater sensor network with depth adjustment capabilities. Through simulations and in-situ experiments, we verify the functionality and performance of the system and algorithm.

Traditional configurations for mounting Temperature−Humidity (TH) sensors on multirotor Unmanned Aerial Systems (UASs) often suffer from insufficient radiation shielding, exposure to mixed and turbulent air from propellers, and inconsistent aspiration while situated in the wake of the UAS. Descent profiles using traditional methods are unreliable (when compared to an ascent profile) due to the turbulent mixing of air by the UAS while descending into that flow field. Consequently, atmospheric boundary layer profiles that rely on such configurations are bias-prone and unreliable in certain flight patterns (such as descent). This article describes and evaluates a novel sensor housing designed to shield airborne sensors from artificial heat sources and artificial wet-bulbing while pulling air from outside the rotor wash influence. The housing is mounted above the propellers to exploit the rotor-induced pressure deficits that passively induce a high-speed laminar airflow to aspirate the sensor consistently. Our design is modular, accommodates a variety of other sensors, and would be compatible with a wide range of commercially available multirotors. Extensive flight tests conducted at altitudes up to 500 m Above Ground Level (AGL) show that the housing facilitates reliable measurements of the boundary layer phenomena and is invariant in orientation to the ambient wind, even at high vertical/horizontal speeds (up to 5 m/s) for the UAS. A low standard deviation of errors shows a good agreement between the ascent and descent profiles and proves our unique design is reliable for various UAS missions.

Small unmanned aircraft systems (sUAS) are rapidly transforming atmospheric research. With the advancement of the development and application of these systems, improving knowledge of best practices for accurate measurement is critical for achieving scientific goals. We present results from an intercomparison of atmospheric measurement data from the Lower Atmospheric Process Studies at Elevation—a Remotely piloted Aircraft Team Experiment (LAPSE-RATE) field campaign. We evaluate a total of 38 individual sUAS with 23 unique sensor and platform configurations using a meteorological tower for reference measurements. We assess precision, bias, and time response of sUAS measurements of temperature, humidity, pressure, wind speed, and wind direction. Most sUAS measurements show broad agreement with the reference, particularly temperature and wind speed, with mean value differences of 1.6 ± 2.6 ∘ C and 0.22 ± 0.59 m/s for all sUAS, respectively. sUAS platform and sensor configurations were found to contribute significantly to measurement accuracy. Sensor configurations, which included proper aspiration and radiation shielding of sensors, were found to provide the most accurate thermodynamic measurements (temperature and relative humidity), whereas sonic anemometers on multirotor platforms provided the most accurate wind measurements (horizontal speed and direction). We contribute both a characterization and assessment of sUAS for measuring atmospheric parameters, and identify important challenges and opportunities for improving scientific measurements with sUAS.

Recharging sensor networks using Unmanned Aerial Vehicles (UAVs) provides a possible method for increasing network lifetime. In this paper, we evaluate that approach, determining how much of a benefit it provides and under what conditions. We base our simulations and field experiments on data collected from charging with our UAV-based wireless power transfer system, which has similar transfer ranges and efficiencies as other such systems. We determine that a UAV can increase the network lifetime up to 290% compared to no recharging, that the UAV should recharge 30% of the sensor node battery capacity at one time for the maximum benefit, and that the UAV should recharge the lowest powered node until the network reaches a size of approximately 306 nodes at which point it should recharge the sink. We also examine how the sensor network can aid this through sink selection. The policy varies as network size increases, with a static approach working well until 200 nodes, and then either a perimeter or heuristic approach works best. These results inform future use of UAVs in recharging and working with sensor networks.

When extending the life of Wireless Rechargeable Sensor Networks (WRSN), one challenge is charging networks as they grow larger. Overcoming this limitation will render a WRSN more practical and highly adaptable to growth in the real world. Most charging algorithms require a priori full knowledge of sensor nodes’ power levels in order to determine the nodes that require charging. In this work, we present a probabilistic algorithm that extends the life of scalable WRSN without a priori power knowledge and without full network exploration. We develop a probability bound on the power level of the sensor nodes and utilize this bound to make decisions while exploring a WRSN. We verify the algorithm by simulating a wireless power transfer unmanned aerial vehicle, and charging a WRSN to extend its life. Our results show that, without knowledge, our proposed algorithm extends the life of a WRSN on average 90% of what an optimal full knowledge algorithm can achieve. This means that the charging robot does not need to explore the whole network, which enables the scaling of WRSN. We analyze the impact of network parameters on our algorithm and show that it is insensitive to a large range of parameter values.