ICUAS 2024 UAV Competition

The main repository for ICUAS'24 UAV competition for submission by ariitk

Running the Code

git clone [email protected]:AerialRobotics-IITK/icuas-2024 && cd icuas-2024 
./docker_build.sh && ./docker_run.sh 
./startup/challenge/start.sh

---

Development on Docker

Installing Docker

Check the official docker engine installation guide

Building and Running Docker Image

Run the following commands in order

• ./docker_build.sh --build-args "--no-cache --pull" --focal-nogpu
  

  --no-cache will ensure that you do not store any cache for any of the docker commands executed other than cached images, hence everytime you rebuild using this command your current working directory (i.e. the package with the latest changes on local) will be copied onto the docker container

  Alternative Command:

  ./docker_build.sh --build-args "--pull --build-arg CACHEBUST=0" --focal-nogpu
  

  CACHEBUST when set to 0 would ensure all the docker commands are run from cache until the value of the argument remains same. In our case, I have modified the Dockerfile for this argument changes right at the command where we don't want data from the cache to be used. This saves some bit of time by skipping system wide installations (once they are atleast run once).

• ./docker_run.sh --run-args "--rm" --focal-nogpu
  

  --rm will ensure that the docker container is always stopped and removed, whenever you all instances of the container have been closed

Removing cached Docker data

If you had run docker without the --no-cache flag, you would have a cached copy of the package. So, any local changes in the workspace will not be reflected on docker no matter how many times you stop/rerun the container. To circumvent around this, you need to remove the cached docker images and files completely and, then use the above two commands for your workflow.

docker system prune --all

---

Problem Statement Details

Rulebook

Check the rulebook.

Simulation

UAV simulation template startup. Tmux session is running on the left side, with the Gazebo client positioned on the right.

Controlling the UAV

For your reference, we have set up trajectory planning using TOPP-RA, which you can use by publishing two topics:

• tracker/input_pose - Send a waypoint (PoseStamped) to TOPP-RA. TOPP-RA then interpolates trajectory between the current UAV pose and the target waypoint and sends trajectory points (MultiDOFJointTrajectoryPoint) to topic position_hold/trajectory with a given rate. The position controller of the UAV receives the trajectory point as a reference and commands the motors.
• tracker/input_trajectory - Generate a trajectory using the given sampled path in the form of waypoints (MultiDOFJointTrajectory). TOPP-RA then interpolates trajectory from the current UAV pose to the first point of the trajectory and interpolates trajectory between sampled path points. Once the trajectory is interpolated, each trajectory point is sent as a reference to the position controller via the position_hold/trajectory topic

To control the UAV directly, and to publish the trajectory that you generated via your solution, you need to use the following topic:

• position_hold/trajectory - Publish a trajectory point directly to the UAV position control

Current position reference (the last one sent to the position controller of the UAV) can be obtained via carrot/pose topic, while the current pose of the UAV (in simulation) is available at odometry topic.

Configuration

Configuration files are placed in the startup/challenge/custom_config folder.

• Position Control
• TOPP Trajectory Generation

Plant beds, plants, and fruits

The goal of the UAV is to count the total number of fruits of a given plant. The list of beds that the UAV needs to visit will be provided on the topic plant_beds within the UAV namespace. The message is a string (std_msgs/String) and contains the name of the plant and beds that need to be visited (eg. Pepper 1 2 8 12 19), separated by whitespace. The number of beds may vary in a given mission.

Plant beds layout.

Plants are simulated via textures on flat surfaces. The fruits of one variety of plants have the same color and shape and are simulated as 3D objects partially visible from either one or both sides of the plant. The shapes and colors of fruits on different variety of plants are different. There are three plant varieties: a) Tomato (red), b) Pepper (yellow), and c) Eggplant (purple).

Challenge

More details on the challenge can be found in the competition rulebook. After the UAV and the world is spawned, plants will also be spawned in beds. One plant bed can contain up to three plants. Each plant may have a different number of fruits (including zero). Some plant beds in the world may be empty, but the ones listed in the mission specification through the plants_beds topic will have at least one plant.

UAV in the simulation arena for the ICUAS 2024 challenge.

The setup in this repo reflects how the world for evaluation will be constructed. Teams are welcome to make a permanent world for testing and development since spawning of a large number of plants in Gazebo takes some time.

Important node for running your code

Remember that you should run your code alongside the existing setup through the session file. Use the existing solution section. You can use either existing waits to spawn your nodes or set your nodes to listen to the following topic:

• challenge_started - After True is published on this topic the UAV is set up correctly.
• plant_beds - After the array of points of interest is published on this topic the arena is set up correctly.

Make sure to wait for data on both topics to safely run your code.

Changing the plant locations for testing

You can change plant locations through the session file or spawn them from a separate terminal. Use position args to change the location of the plant.

Sign convention used and Distances between beds

Green: Y Axis Red: X Axis

• Gap between shelf: 6.00 ; Range(4.00, 16.00)
• Gap between shelf: 1.50 ; Range(4.5, 22.5)
• Gap between z levels : 2.80 ; Range(1.1, 6.99999999999999)