Skip to content

Latest commit

 

History

History
97 lines (66 loc) · 3.38 KB

README.md

File metadata and controls

97 lines (66 loc) · 3.38 KB

Row Crop Follow

Nodes

seg_controller_node

This node can run the controllers SegZeros, SegMin and SegMinD described in the paper GPS-free Autonomous Navigation in Cluttered Tree Rows with Deep Semantic Segmentation.

Topics

Set topic names through their parameter

  • Subscriptions:
    • rgb_topic
    • depth_topic
  • Publishers:
    • cmd_vel

Parameters

controller_type: SegZeros, SegMin or SegMinD

cmd_vel_topic, rgb_topic, depth_topic

model_file: full path to the model file

inference_device: could be either cpu or gpu. For TFLite models cpu is required.

normalization: normalization used for DNN input preprocessing. Default: imagenet

input_width_network, input_height_network

lin_vel_gain, ang_vel_gain

max_lin_vel, max_ang_vel

ema_filter_alpha: exponential factor used in the Exponential Moving Average filter applied to smooth the velocities in output. Set to 1.0 to disable.

depth_threshold: maximum depth value in meters. Values over this threshold are saturated to this value.

confidence_threshold: confidence threshold for the prediction of the DNN

show_result, publish_processed_image

otsu_thresholding_node

This is an implementation of the work "Autonomous Navigation and Crop Row Detection in Vineyards Using Machine Vision with 2D Camera"1 used for comparison.

Topics

Set topic names through their parameter

  • Subscriptions:
    • rgb_topic
  • Publishers:
    • cmd_vel
    • ~/result_img

Parameters

cmd_vel_topic, rgb_topic

propotional_criterion, deviation_control_gain, correction_control_gain: respectively PC, G and P as described in the original paper

linear_velocity

grayscale_method: Choices: cv2, 2gbr, gb. The two althernative methods described in the paper, 2GBR and GB, where both implemented and can be used. Additionally, it is provided the possibility of using the classical conversion provided by OpenCV.

show_result: whether to show the result of the control algorithm using OpenCV.

Usage

Requirements:

  • ROS2 Humble
  • TensorFlow 2.x (2.10 tested)

Installation

  • Clone the repo in your workspace with git clone https://github.com/PIC4SeR/SegMinNavigation.git
  • Install the required dependencies with rosdep install --from-path src --ignore-src -y
  • Build the workspace with colcon build --symlink-install --packages-select row_crop_follow

Run the application

Two launch files are provided for convenience and can be found in the launch folder. For both launches a parameter file can be set as launch argument using params_file:=path-to-file.

Example:

 ros2 launch row_crop_follow seg_controller.launch.py params_file:=~/my_params.yaml

Footnotes

  1. Mendez, E.; Piña Camacho, J.; Escobedo Cabello, J.A.; Gómez-Espinosa, A. Autonomous Navigation and Crop Row Detection in Vineyards Using Machine Vision with 2D Camera. Automation 2023, 4, 309-326. https://doi.org/10.3390/automation4040018 ↩