Status

Work-in-progress

Bief

Adaptive object detection, and obstacle avoidance robot using a Jetson Nano 2GB, OpenCV, Lidar, OpenCR and ROS. A ROBOTIS TurtleBot3 waffle is used as a hardware platform starting point. A Jetson Nano 2GB will be used instead of the included Raspberry Pi 3B+ to enable potential of GPGPU use in upcoming iterations of the robot.

Components

Hardware:

1. JetsonNano 2GB
2. ROBOTIS TurtleBot3 waffle kit, including:
   1. LDS-001 Lidar sensor + USB2LDS interface board
   2. OpenCR v1.0
   3. 2x servo motors
   4. RPi camera v2.1 + CSI ribbon cable
   5. Battery Pack/power supply.

Software:

1. JetPack v4.6.1, which includes:
   1. OpenCV v4.1.1
2. ROS Melodic Morenia <!– 3.
6. –>

Demo

• Robot, up and running with a Jetson Nano.

• Robot’s POV of the scene ahead via camera stream as-is, and with yaw adjustment based on centeroid calculations.

• Experimenting with different color spaces using basic color masking and a bitwise_and operator on each pixel to isolate the object of interest.

  • Absolute red mask (RGB mask = (255, 0, 0)) with a threshold tolareance of +/- 40 intesity levels in each channel.

  • Absolute green mask (RGB mask = (0, 255, 0)) with a threshold tolareance of +/- 40 intesity levels in each channel.

  • Absolute yellow mask (RGB mask = (255, 255, 0)) with a threshold tolareance of +/- 40 intesity levels in each channel.

• x

Design

Software block diagram

Software statemachine flow diagram

Future work

Imageg processing

1. Implment an ambient light sensor (e.g., TCS34725 to enable dynamic threshold adjustment (consider V channel in HSV color space).
2. Implment active learning model (consider TensorRT+CUDA), and compare accuracy and performance to other iterations of this project.

Motion Control

1. Design and implment a state-space PID controller model to effeciently navigate with yaw adjustment in mind concurrently.