This project implements a hand gesture-controlled vehicle system using fuzzy logic.
The car's movement is controlled by hand angles captured from the user's gestures.

• Start Detection: Show a thumbs-up gesture to start gesture control.
• Stop Detection: Open your palm fully facing the camera to stop gesture control.

• Turning the right hand to the right (positive angle) makes the car turn right.
• Turning the right hand to the left (negative angle) makes the car turn left.

• Holding the left hand vertically sets the car to the lowest speed.
• Tilting the left hand to either side increases the speed.

• Implements the fuzzy control system
• Input: Right and left hand angles
• Output: Car's steering angle and speed

• Implements the linear control system

• Draws the car and visualizes its movement on a 2D plane.

Installing Mediapipe 0.8.5 on jetson Nano

• We follow this link to install mediapipe-GPU on Jetson[https://jetson-docs.com/libraries/mediapipe/overview]
• To build the mediapipe-CPU
  • First, you need to install bazel in your system specifically 3.7.2 for 0.8.5
  • Clone this mediapipe files [https://github.com/google-ai-edge/mediapipe]: git clone -b 0.8.5 https://github.com/google-ai-edge/mediapipe.git
  • Run the command: cd mediapipe
  • Run this command to build the C++ program python3 setup.py install
  • Then create the wheel python3 setup.py bdist_wheel
  • Now you have folder named dist: pip install dist/filename.whl

To launch the gesture-controlled car simulator, run:

python hand_control_sim.py --input camera # Using camera
python hand_control_sim.py --input video --video_path inputs/short_straight.mp4 --controller linear # Using pre-record video

To launch the gesture-controlled remotely, run:

python server_robot_receiver.py # Robot side
python client_camera_sender.py # Client side, need to modify the IP to server's IP

Evaluate GPU efficiency on the Jetson Nano:

python3 hand_control_efficiency.py

Evaluate CPU efficiency on the Jetson Nano:

python3 hand_control_CPU.py