AutoDogFeeder

An automatic dog feeder based on ROS 2(iron) running on a RPi4 B.

Instructions to Run

• On the Raspberry Pi, grab the docker image by running

  docker pull n33r4j/auto_dog_feeder:v0.1

  Note: This might take a while.

• Clone this repo in a convenient location.

• Assuming docker is installed on the RPi, run the container with

  docker run -it -v ~/<path to cloned repo>/AutoDogFeeder:/AutoDogFeeder \
  -v /dev/video0:/dev/video0 \
  --device-cgroup-rule='c 81:* rmw' \
  n33r4j/auto_dog_feeder:v0.1
  

• Update dependencies with

  cd ~/AutoDogFeeder

  rosdep install -yr --from-path src

  (might be better to put this in an entrypoint.sh or other script.)

• Build the package and source local workspace.

  colcon build

  source install/local_setup.bash

• Run the launch file. TODO

Alternatively, if you have the environment already setup(without docker?)

• Clone repo
• cd AutoDogFeeder/
• chmod +x run_demo.sh
• ./run_demo.sh which runs the Camera node.
• You can trigger video recording by switching between publishing a 0 and 1 to the /camera_mode topic in a new terminal with,

ros2 topic pub -r 10 /camera_mode std msgs/msg/Int8 "{data: 1}"

Assumptions

• Only works for small dogs(as in the specs).
• No other pets in the household and minimal other disturbances that might trigger the feeder.
• Powered from a wall outlet(no batteries).
• Gravity based food and water dispensing with minimal jamming.
• Food and water dispensing is coupled.
• The dispenser servo reaches a target in negligible time.
• The dispenser is designed in a way to achieve controlled flow of water and food, i.e. something like a tap.
• The dog is not scared of the automatic feeder or the sounds it makes.

Systems Diagram

• Raspberry Pi 4
• USB Camera
• PIR Sensor
• Mini servo motor
• LCD Display
• Speaker

Considerations

Dog Detection Method:

Several methods are possible depending on factors like size of the dog,no. of dogs(or other pets), possible disturbances, and available compute resources.

1. manual trigger that the dog can be trained to hit(activates near feeding time)
2. camera-based motion detection
3. other sensors (ultrasonic, IR etc)
4. RFID collar/ QR code collar
5. YOLO or pet facial detection

Camera Placement/Resolution

Depends on the exact purpose of the camera recordings.

• track dog feeding behavior?
• track leftover food/water on dish?
• allow owner to see/interact with dog remotely?

Video Storage:

• Overwrite old recordings(default)?
• Backup to a local server/cloud?
• Compress videos?

Feeder:

• You may need to dispense water more often than food.
• Might need to preempt/retry the door mechanism if jamming is an issue(unless the design prevents this). Need to choose between complicating software or hardware.

Testing

• Tests for each node using pytest?
• Create dummy nodes that publish sensor values, call services/actions.