Because of this, it is impossible to make this component naturally degradable by the current mode of operation. Fortunately, the mass of this component is minimal – only about 0.5 mg. To put that into perspective, it would take 10,000 ion-selective membranes to produce as much plastic pollution as a single credit card. Table 5.3 shows all of the components in a wireless nitrate sensor node and what naturally degradable materials they can be substituted with.Some components of a conventional wireless sensor node are difficult or even impossible to replace with naturally-degradable materials, as shown in Table 5.3. For example, degradable batteries or other energy storage devices exist in literature, but none are resilient or low-cost enough for our application. Similarly, using onboard energy storage and harvesting necessitates a higher complexity micro-controller, which corresponds to larger and more costly micro-controllers. One method of circumnavigating these components is using passive sensor nodes, such as passive RFID sensors. Passive RFID sensors comprise an antenna, an RFID IC, and a sensor. Of note, there is no onboard energy storage, meaning an external power signal must be sent to the node to take a measurement. In the case of RFID, an RF signal is transmitted by an external RFID reader. The antenna receives the wave and transduces it into an electric signal which ‘wakes up’ and powers the RFID IC. The RFID IC acts as the micro-controller, communications IC, and power management. When it receives the wake-up signal, it uses the power in that signal to read the sensor and modulate a return signal through the antenna to the reader corresponding to the sensor measurement. By designing a sensor node using this passive RFID scheme, we estimated that we can make the naturally-degradable nitrate sensor nodes 99.99% degradable by mass. Ag/AgCl strips were fabricated using the same parameters with Engineered Materials Systems, Inc. CI-4001 ink. Afterward, they were cured in an oven at 120◦C for two hours. After curing, the carbon and Ag/AgCl strips were cut into six equal-sized electrodes. Each electrode was then sandwiched between two patterned wax sheets and heated in an oven at 55◦C for thirty minutes. The wax sheets were made by soaking untreated plywood sheets in water before dipping them in molten wax and removing the waxy film that forms on the surface. The thin water layer on the surface of the saturated plywood sheet acts as a barrier to the hydrophobic wax, allowing for easy removal. The thickness of the wax sheets was controlled by dipping the saturated plywood sheets multiple times in quick succession, obtaining wax sheet thicknesses of 350 µm, 700 µm, and 1.25 mm for one, two, and three dip cycles, respectively.

The wax sheets used for encapsulating the bottom of the sensors were used as-is, while the sheets used for encapsulating the top of the sensors had 12.5 µm windows for the membranes removed using a laser cutter. An image of an ISE immediately after the encapsulation step is shown in Figure 5.17B. ISE membranes were fabricated by mixing 5.2 wt% Nitrate Ionophore VI, 47.1 wt% dibutyl phthalate, 0.6 wt% tetaroctylammonium chloride, and 47.1 wt% PVC. A total of 0.2 g of this mixture was dissolved in 1.3 mL of THF. 180 µL of the membrane solution was drop-cast on the ISE surface and dried in a fume hood for 15 minutes. The REs employed a CNT transducer layer between the Ag/AgCl electrode and the membrane. This transducer was composed of 0.01 g of CNT and 0.05 g of F127 -block-poly-block-poly diacrylate dissolved in 10 mL of THF, which were sonified for 1 hour in an ice bath using a Branson Digital Sonifier probe. 120 µL of the resulting transducer cocktail was deposited onto the RE surface. The salt membrane was made by dissolving 1.58 g of Butvar B-98 , 1.00 g of NaCl, and 1.00 g of NaNO3 in 20 mL of methanol. The mixture was sonified for 30 minutes in an ice bath, and 180 µL of the resulting salt membrane cocktail was deposited on top of the CNT transducer. Unless otherwise noted, all chemicals used in ISE and salt membranes were obtained from Millipore Sigma. After each electrode was made, they were cold-sintered to 22 AWG wire using 8331D silver conductive epoxy and en-capsulated with multiple layers of Gorilla 2-part Epoxy . Figure 5.17 shows an image of the fabricated naturally-degradable nitrate sensors. Wireless sensor networks are becoming more and more relevant in agriculture. Researchers have made agricultural WSNs to monitor weeds, evapotranspiration, crop disease, and water use. However, there are limited examples of agricultural WSNs for monitoring nitrate. The design of a wireless sensor network in agriculture has a host of unique challenges. Issues like energy consumption for autonomous operation of sensor nodes dictate design and development issues, including communication protocols and deployment. Furthermore, the placement of sensor nodes in open, uncontrolled environments presents another host of unique challenges, such as damage accumulation from weather or wildlife. Finally,blackberry plant pot the scale it takes to implement WSNs in agricultural settings is much larger than in commercial or industrial environments. Cropland accounts for about 11% of the habitable land globally, and in the United States, the average crop farm is 445 acres. This dictates the placement and quantity of sensors needed, as discussed in Section 5.2, and shows that large numbers of sensor nodes are required. Different researchers have adopted different strategies for circumventing these challenges. Ding and Chandra investigated using Wi-Fi for measuring soil moisture and electrical conductivity.

Syrovy et. al. utilized Long Range, Wide Area Network communications to transmit data from paper-based soil moisture sensors. Yu et al. deployed a system where the sensors connect directly to a person’s phone over Bluetooth Low Energy. Here, we propose an agricultural WSN explicitly designed for the precision management of soil moisture and soil nitrate. The naturally-degradable nitrate sensor nodes demonstrated in Section 5.4 can be deployed at minimal cost and without the need for maintenance throughout any agricultural field using the techniques outlined in Section 5.2.Hence, a reader needs to be brought to within a few meters of the sensor to sample data from the sensors. Because many sensors need to be distributed across an agricultural field to acquire granular enough data to capture soil variability, drones offer a unique advantage over other existing methods to sample data from the sensors. With drones and drone accessories becoming less expensive, using multiple drones to simultaneously map sensors has become an attractive route to efficiently gather data. Machine-learning algorithms are a promising approach for generating flight path maps due to their ability to solve highly non-convex problems quickly, and even operate in real-time as a digital twin. We developed an agent-based dynamics model to generate flight paths for the drones to scan each sensor in the field while circumventing obstacles and avoiding crashes. The coordinated effort of multiple drones working towards a common objective has similarities to swarms found in nature, such as bees and ants, where the accumulation of each agent’s actions and reactions can give rise to phenomena and emergent behavior where the system becomes more than the sum of its parts. Unlike bees and ants, it is atypical for a drone swarm to contain a ‘leader.’ In the context of field mapping, the drone swarm adapts to changes within the system, such as the disablement of a few drones due to collisions or other unforeseen causes. We developed a robust agent-based model capable of optimizing the flight paths of each drone within a swarm to scan all sensors within a simulated agriculture field. The simulations determine each drone’s aerial route for optimal flight path planning. Each drone within the simulated framework – an ‘agent’ – has its own characteristics that determine how it interacts with its surroundings, such as its environment and other drones. These characteristic parameters take inspiration from the physics of molecular dynamics, where each agent is modeled as a point mass particle that is attracted and repelled by other objects within the system. A genetic algorithm determines the direction of propulsion. The framework inputs are the field’s shape, the number of agents, and the positions of sensors . This framework can be used for various sizes and shapes of agriculture fields. Depending on the field geometry and the locations of sensors within that field, the framework will output several suggestions of each drone’s flight path trajectory. AUTONOMOUS agricultural mobile robots become increasingly more capable for persistent missions like monitoring crop health and sampling specimens across extended spatio-temporal scales to enhance efficiency and productivity in precision agriculture. An autonomous robot needs to perform certain tasks in distinct locations of the environment subject to a specific budget on the actions the robot can take . During in-field operations, the actual costs to complete tasks can be uncertain whereas expected costs may be known. Also, some tasks can be more urgent than others, and have to be prioritized. It is often the case that there exists some prior information about a required task that can bias robot task assignment. Hence, it is necessary to develop approaches that utilize limited prior information to plan tasks with uncertain costs and priority level. There exist two key challenges for efficient robot task allocation in precision agriculture. First, prior maps can indicate biases in task assignments, but may not be trustworthy. This is because conditions in the agricultural field can change rapidly, are dynamic, and may be hard to predict ahead of time. Second, as the budget is being depleted, the robot needs to periodically return to a base station . Addressing these two challenges simultaneously poses a two-layer intertwined decision making under uncertainty problem: How to perform optimal sampling given an approximate prior map, and how to decide an optimal stopping timeto avoid exceeding a given task capacity? This paper introduces a new stochastic task allocation algorithm to balance optimal sampling and optimal stopping when task costs are uncertain. A direct approach for persistent sampling is to survey the entire space and perform the desired task sequentially. The main drawback is that the robot would then exhaustively visit all sampling locations without prioritizing those that would yield a higher gain or would be more time-critical. Orienteering can address part of this drawback by determining paths that maximize the cumulative gain under a constant budget. The robot prioritizes visiting adjacent locations if they jointly yield higher gains than isolated high-gain locations, and provided that any budget constraints are not violated. However, this strategy can be insufficient for missions where some tasks are more urgent than others. For instance, several existing robot task allocation strategies, albeit for distinct application domains, typically consider a deadline or user-defined importance levels.