VIMAN - VIrtual Manual and Autonomous Navigation

With Project VIMAN, we aim to build a quadcopter capable of maneuvering in autonomous mode with the ability to detect people having high body temperature in a crowded area. This capability could be used to churn out potential CoVID-19 infected people from a crowd.

In order to incorporate autonomy into our UAV, the following milestones are to be achieved:

• • Control via keyboard (Stage 0)
• • SLAM Z (Stage 1)
  • • Hover at a set height (Stage 1.1)
  • • Hover with a set heading (Stage 1.2)
  • • Linear-Z mapping (Stage 1.3)
    • • Camera calibration (1.3.0)
    • • Color thresholding (1.3.1)
    • • Color identification (1.3.2)
  • • Rotation-Z mapping (Stage 1.4)
  • • Z mapping (Stage 1.5)
• • SLAM X (Stage 2)
  • • Detect motion in X direction using camera (Stage 2.1)
• • SLAM Y (Stage 3)
• • Complete SLAM (Stage 4)

Note: Addition/deletion of sub-stages is dependent on the complexity of implementation of a stage.

List of sensors used

1. Altimeter
2. IMU
3. Camera

About the repository

This repository contains two ROS packages:

1. viman_control: Contains plugins to simulate the quadcopter, sensors, and ROS nodes to test and process vision and implement control algorithms.
2. viman_visualize: Contains 3D models that build a Gazebo environment and the model of the quadcopter itself.
3. viman_utility: Contains utility nodes and custom messages to ease the development and control of VIMAN.

Note: Make sure you have ROS' cv_bridge because OpenCV 3.2 with Python2.7 has been used for vision processing.

Implementation

0 | Control via keyboard

Step 1: Execute the following command in terminal to launch the Gazebo world with viman.

roslaunch viman_visualize gazebo-disp.launch on_rviz:=true

Step 2: Open another terminal and execute the following command to use keyboard keys to control the quadcopter.

rosrun viman_control viman_key_ctrl

Step 3: Play the simulation in Gazebo, place the focus of the terminal opened in step 2 and read the instructions provided by the ROS node.

---

1 | SLAM Z

1.1 & 1.2 | Hover at a height with a set heading

Step 1: Execute the following command in terminal to launch the Gazebo world with viman.

roslaunch viman_visualize gazebo-disp.launch on_rviz:=true

Step 2: Open another terminal and execute the following command to use keyboard keys to control the quadcopter.

rosrun viman_control viman_sa

Step 3: Play the simulation in Gazebo, place the focus of the terminal opened in step 2 and read the instructions provided by the ROS node.

1.3 | Map the features observed by VIMAN and store it into a file

Step 1: Execute the following command in terminal to launch the Gazebo world with viman.

roslaunch viman_visualize gazebo-disp.launch nav:=1

Step 2: Open another terminal and then execute the following command

rosrun viman_control vm_z_linear_nav

Follow the instructions preseted by the ROS node.

1 | Map the features observed by VIMAN and optimize the graph

Perform step 1 from 1.3 and then execute the following command in another terminal.

rosrun viman_control vm_z

Follow the instructions preseted by the ROS node.

Gazebo World

The following is an image of the stack of cylinders that the UAV must map. The shown stack is used to complete stage 1.3.

The following GIF shows the partial cylinder stack (spawned via nav:=2. The environment shown is used for SLAM Z)

Utility

To see sensor data (IMU and altimeter)

rosrun viman_control vm_sensor_data

To view camera output (SLAM Z)

rosrun viman_control z_vision.py

To view the color identified by the camera (SLAM Z)

rostopic echo /viman/color_id

To visualize optimized and unoptimized maps (works only with vm_z)

rosrun viman_control visualize_map

Once the ROS node is run, it shows two plots:

1. Unoptimized map
2. Optimized map
   

To find HSV color ranges for color thresholding

Make use the following changes in frnt_vision.py

...
from vision_process import DisplayImg, Output # For HSV thresholding
...
...
...
th_processing = DisplayImg() # For HSV thresholding
...
...

roslaunch parameters

Note: Parameter values in bold correspond to default value.

1. gazebo-disp.launch:

• on_rviz: false/true | To show or not to show RViz displaying the front camera output.
• nav: 0/1/2 | Choose type of navigation:
  • 0 - Empty world
  • 1 - Additional stacked cylinders for only z linear navigation
  • 2 - Parital cylinders stacked for SLAM Z navigation

Other VIMAN repositories

1. viman_vo : Repo containing python code related to visual odometry.
2. Viman-Dead_Reckoning : Repo containing cpp code to obtain odometry via IMU.

References

• sjtu-drone repository for simulation base.
• This answer from stack-overflow for getch equivalent in Ubuntu.
• dbscan-cpp repository for implementing DBSCAN algorithm to optimize map.
• matplotlib-cpp repository for visualize_map ROS node.
• Google search