Interns: Ansel KurienUG, Cy WestbrookUG
Advisors: Aggelos Bletsas, Richard Howard
Developing intelligent backscatter tags and an automated multimodal plant-sensing platform using differential temperature, NDVI, and passive/active ultrasonics to detect hydration stress.
This Week's Work:
During our first week at WINLAB, we split the project into plant sensing and direction-of-arrival sensing. For plant sensing, we reviewed ways to track plant health through temperature, optical measurements, and ultrasonic activity.
We also studied how backscatter tags could reduce the hardware needed for direction-of-arrival sensing. Our next steps were to test the MSP430 and EFR32FG23 microcontrollers.
This Week's Work:
During our second week at WINLAB, we began designing the power sections of the backscatter tag. This included the circuitry around the EFR32FG23 microcontroller and a board power system based on a solar cell, rechargeable battery, and voltage regulator.
We also wrote MSP430 C code to generate square PWM waves and used MATLAB to select tag frequencies for a multistatic network. For plant sensing, we began reproducing earlier differential temperature results.
This Week's Work:
During our third week at WINLAB, we continued work on the wireless sensor hub that will collect and transmit data from the plant experiments. We tested existing wireless hardware while continuing the EFR32FG23 circuit design.
We also set up an ultrasonic interferometer to study how ultrasonic signals change around the plant. For differential temperature sensing, we assembled the plant enclosure and positioned the sensors needed to compare leaf temperature with the surrounding environment.
These setups formed the base for combining differential temperature, ultrasonic sensing, and later optical measurements into one plant-monitoring system.
This Week's Work:
During our fourth week at WINLAB, we added NDVI color sensing to the plant setup. The circuit used optical measurements to compare how the leaf reflected different wavelengths of light. This gave us another way to track changes in plant condition.
We also built an ultrasonic frequency response setup with two transducers placed on opposite sides of a leaf. Using the oscilloscope, we began measuring how the plant affected the signal across different frequencies.
For differential temperature sensing, we moved the plant into a reflective enclosure so lighting and airflow would remain more consistent. We also tested the ultrasonic microphone and continued working on the frequency generator used for backscatter experiments.
This Week's Work:
During our fifth week at WINLAB, we brought the plant sensors together inside one controlled chamber. The setup included the NDVI sensor and the other hardware needed for repeated plant measurements. This gave us a more stable environment for comparing sensor readings over time.
We also worked on the RF section of the backscatter tag. We selected component values for impedance-matching networks designed around the 868, 915, and 920 MHz bands. Separate designs were considered for different output power levels.
The PCB work also required us to compare component sizes and plan the RF traces. We began routing microstrip transmission lines so the board could maintain the correct impedance between the radio and antenna.
This Week's Work:
During our sixth week at WINLAB, the wireless sensor hub PCB reached the final stages of design. The board combined the EFR32FG23 microcontroller with its power circuitry and RF hardware. It also included connections for the solar panel, battery, and plant sensing modules.
We reviewed the front and back of the board through 3D renders. This helped us check component spacing, connector access, and trace placement before preparing the design for fabrication. Plant sensing tests also continued inside the integrated chamber while the PCB work moved toward completion.
This Week's Work:
During our seventh week at WINLAB, we sent the wireless sensor hub PCB for fabrication and ordered the components needed for assembly. This marked the transition from board design to preparing the physical hardware for testing.
We also developed tools for analyzing passive ultrasonic recordings. The recordings were separated into frequency bands up to 80 kHz so we could locate peaks that may represent plant-generated ultrasonic activity.
For the full plant experiment, we combined active ultrasonic Bode measurements with NDVI and pip tag data. The plots were aligned with the pump state so changes in the sensor readings could be compared with watering and drying periods.
This Week's Work:
During our eighth week at WINLAB, we tested the wireless sensing hardware and worked on communication between the development boards.
We also processed the active ultrasonic measurements collected across several watering cycles. The individual Bode plots were combined into gain and phase spectrograms so changes could be tracked over time and compared with when water entered or left the plant setup.
For passive ultrasonic sensing, we developed an event detection pipeline for the overnight recordings. It identified possible plant clicks while separating them from audible sounds and narrowband electronic noise. This gave us a cleaner set of events to compare with the plant’s hydration state.