Project
# | Title | Team Members | TA | Documents | Sponsor |
---|---|---|---|---|---|
42 | Cat Collar |
Ching Chieh Yang Junnun Safoan Taha Anwar |
Dongwei Shi | design_document2.pdf final_paper1.pdf presentation1.pptx proposal1.pdf |
|
Group: Taha Anwar (tanwar2), Junnun Safoan (safoan2), Ching Chieh Yang (cyang87) Introduction: This is a sponsored project by Petronics, a company that builds Mousr, a mouse robot that plays with cats. We aim to create an IMU-based cat collar that measures, transmits and analyzes the cat’s activity when playing with a Mousr robot. When it is detected that the cat is playing with mousr, a camera is turned on to monitor the cat. Problem: Cat owners do not always have time to play with their cats. The Mousr is a clever solution that accompanies your cat in your absence. However, Petronics still doesn’t have a way to check if there is a direct correlation between a cat’s overall activity and Mousr’s activity. Solution: The Mousr unit developed by Petronics is already able to make event predictions such as ‘inactive’,’engaged’,‘needs charging’ and so on. The motivation for the cat collar is to use the data collected from it to confirm the event predictions by the Mousr in order to assess the cat’s actual engagement with Mousr. Furthermore, the collar will wake up the Mousr when the cat is within a certain vicinity. This will be instrumental for Petronics, as it will allow them to measure the effectiveness of Mousr in engaging with the cat. The collar tracks the motion of the cat using IMU sensors. The data from the sensors is compressed from 120-200Hz to 1 Hz using signal processing for efficient data acquisition. We will develop algorithms to determine whether the cat is: inactive, walking, or engaged with mousr using the data collected. In engaged mode, the raspberry pi turns on a camera for recording the cat’s activity, which serves to verify the results. Mousr also sends events to Raspberry pi, which helps to find out how the Mousr events correlate to the cat’s activity. Examples of Mousr events are running in circles, zigzag, stuck in corner, rest, etc. Hardware: We will be using an ESP32, which is a low-power system-on-chip series with Wi-Fi and bluetooth capabilities for communication with the Mousr, as well as uploading data from the collar using wifi. We will use the camera on the Raspberry Pi to record video data, which will serve as a ground truth for the cat’s activity. We will use the accelerometers and gyroscopes in the IMU to track the motion of the cat. The power supply unit would use batteries that are small enough to be easily integrated with a traditional cat collar and can go on upto 6-8 hrs without charging. Therefore, the main components that will be included in the cat collar are ESP32, IMU, Charging/power circuit, status LEDs and other features we can add on in the future. Software: We will design an algorithm that takes in IMU data from the collar of the cat and Mousr events as inputs to generate an output determining whether the cat is inactive, walking, engaging, etc. Mousr events will be coming in from Python Flask framework and API that the Petronics team has built, while acceleration timestamp values come in as raw data from IMU sensors from the collar. It is expected that inactivity is easiest to predict, since the IMU readings will be mostly constant. We can determine if the cat is walking by checking for spikes in the z-axis and y-axis of accelerometer readings. Engagement is expected to cause rapid changes in the accelerometer readings, however, it will be a challenge to determine different motions during play. When engagement mode is detected, the camera turns on to record the cat’s activity. Our final representation will involve either logging the results to a csv file or developing the front end for displaying a simple chart with percentages of cat’s daily activities as well as correlation between cat activities and mousr events. Criteria For Success: Our criterion of success is to design a collar that can accurately predict when the cat is inactive/sleeping, walking and engaged with the Mousr unit. Reach goals would be to delve deeper into the activity of engagement through experimentation and provide more details about its motions during play. |