WP1: Sensors

Week Progress Update Actions/Analysis
15 Work package completed Sensors combined into RF communication and ESP32 board. The sensors are handled off so this work package is completed
18 Finalised parts for the solar power, battery recharge circuit, and power electronics to power the sensor and components. Ordered the part for power electronics after analysing the options and seeing which option would be best for this application.

WP2: Central Hub

Week Progress Update Actions/Analysis
15 The first iteration of machine learning code was added, and integration of sensor code. The temperature and humidity sensor worked as expected but the NPK sensor still struggled to achieve consistent readings. Altium chosen as PCB design software. This was downloaded and familiarised with. The PCB design is to enable and allow the connections between all of the central hub components. It doesn’t require any SMD components which allows the design to be simpler. The components on the PCB must be layout such that they can be accessed from outside the case. For example the SD card and antenna need to be on the outside.
16 First planning for the recharge circuit for both the sensor and central hub. Both devices need to be portable, therefore requiring a battery and adequate power electronics to power the correct voltage levels to the microcontroller. Luckily some options for recharge/power distribution ICs already exist. These devices allow the battery to provide power to the device but also allow the battery to be charged while the power source power the devices. They also have many safety features built-in. To go with this there also needs to be some voltage regulation on the output as the battery level is not consistent and the ESP32 requires an accurate 3.3V. Finally some microSD card modules are required to enable data storage on the central hub and system.
17 Progress was halted due to the assignment's automatic extension for ES434. N/A
18 Attempted to implement the newest machine learning code into the central hub software but run into problems. Finalised part list for the recharge circuit and parts ordered. PCB design was progressed on Altium, but required some changes based on changes to the power electronics and accessibility requirements for the SD card and RF module. The newest version of the machine learning code uses more complicated data structures which compiles but attempts to access illegal memory at runtime. This problem occurs on both the microcontroller and a computer, therefore, suggesting that the generated C code is not functional. Update PCB designs based on requirements.
19 Newest machine learning model now sucessfully implemented. The plan to go back to the original version of the model and create the new part from scratch in C worked effectively. The power electronic parts were delivered and soldered. The previously generated code was causing memory errors in C. This was due to a change that tried to add an addition layer to the decision tree. Instead this addition layer was written in straight C code, which was fairly straight forward due to it being basic logic steps to follow which decision to choose. After the new parts were delivered, they were soldered and tested. The voltage levels needed to be tested to ensure that they were a suitable level to incorporate in the system. All the parts were tested and worked as intended. The only missing part is the battery which took longer to get approved and delivered.
20 PCB design finish and checked. The PCB only required a few small changes to incorporate the correct voltages inputs based on the newest power electronic design. After a large amount of checking the PCB was deemed to be reading for printing. The PCB for both the central hub and sensor contained many of the same parts, making the decision process smooth to do them both. After figuring out how the PCB design software work making the PCBs was relatively quick. The PCBs are an effective way to decrease the size of the prototype and ensure that all the connections are secure.
21 The PCB job request was filled in and sent to the technician for a final check. The check was approved and the PCB was ordered. While waiting for the PCB to be deliver, free time was spent writing up evidence for the central hub. The job request form was filled in and approved by the technician to get the PCB ordered. Waiting for the PCB to be delivered to solder it and do some testing. As the deadline for the report approaching, the focus needs to switch away from technical work and towards writing the report. Many diagrams were created to illustrate the functionality of the central hub and sensor.
22 Sensor PCB delivered. Worked on final Design Portfolio submission including writing up the Technical Report, Design Narrative and Portfolio of Evidence. Soldered Sensor PCB that was delivered, delivery with Central Hub PCB has been interrupted.
23 Worked on final Design Portfolio submission including writing up the Technical Report, Design Narrative and Portfolio of Evidence. All members worked on these submissions.
24 Worked on final Design Portfolio submission including writing up the Technical Report, Design Narrative and Portfolio of Evidence. All members worked on these submissions.

WP3: Machine Learning

Week Progress Update Actions/Analysis
15 Version 1 of the final training dataset was created using the optimum values for temperature, humidity and NPK values. The training dataset was tested using test data and provided a validation accuracy of 100% (though noise and test data still need to be added). This shows, however, that the training dataset can produce effective decisions from the decision tree algorithm.
16 Generated code implemented into the breadboarded central hub. Initial decision tree MATLAB code was converted to C and implemented into the central hub. Produced the correct output with the current NPK values (dependent on this sensor to test other outputs).
17 Version 2 of the final training dataset was created, including new scenarios for when the conditions were unsuitable for rice growth, such as low humidity and high temperature. In addition, an attempt to quantify how much fertiliser should be added to the soil. The training dataset was edited in Excel to provide specific reasons to the user why the conditions were unsuitable for growth. Also, an ‘IF’ statement was added to the function to modify the ‘Add fertiliser’ outcome.
18 Version 3 of the final training dataset was created, incorporating specific outcomes and advising the users on what fertilise to add and how much of it. We added extra data points to the dataset and edited the labels to the appropriate outcome. An ‘IF’ statement was also implemented to specify to the farmer how much N, K, or P fertilisers to add and at which stage.
19 Generated code was compiled on the microcontroller but did not produce an output when the ‘Add fertiliser’ statement was called due to the added ‘IF’ statements. In order to fix this issue, the code was generated without the additional ‘IF’ statements, and the ‘IF’ statements were implemented directly in the C code rather than implementing them in the generated code. This allowed the microcontroller to process the data more easily and reduce memory usage, and the code managed to upload and work.
20 Wrote up evidence section of design portfolio based on the new training dataset and ‘IF’ statements. Created test cases to test the new model. Updated Notion and wrote up validation accuracies and tested test data validation accuracies. Code was written to test the test cases and calculate the validation accuracies.
21 Started planning the portfolio layout and wrote up design narrative with respect to machine learning section. Condensed the evidence section written on Notion, made flow charts explaining the model requirements and screenshotted evidence for the validation accuracies of the model. We also made a note of the literature we wanted to add in the Literature Review of the Technical Report.
22 Worked on final Design Portfolio submission including writing up the Technical Report, Design Narrative and Portfolio of Evidence. All members worked on these submissions.
23 Worked on final Design Portfolio submission including writing up the Technical Report, Design Narrative and Portfolio of Evidence. All members worked on these submissions.
24 Worked on final Design Portfolio submission including writing up the Technical Report, Design Narrative and Portfolio of Evidence. All members worked on these submissions.

WP-ST: Stretch Targets

Week Progress Update Actions/Analysis
16 Application Idea and planning:
-market research completed
-audience identified
Came up with an initial idea of what features to add the the App. Preformed market research by studying the UI/IX of educational mobile applications. The purpose of the app is to assist in the education of farming techniques for Indian farmers (specialising on fertiliser misuse). A plan for development was created.
17 UX/UI design was decided: