Skip to content

final_project.md

Latest commit

 

History

History
154 lines (119 loc) · 9.68 KB

final_project.md

File metadata and controls

154 lines (119 loc) · 9.68 KB

Final Project

Before You Start

This final project is dependent on HW1 - HW5. We provide you with the solution for HW1- HW3. Please start this homework after you finish HW4 and HW5.

We recommend creating a new folder, finalhomework_ws, for the final project and copying your HW1–HW5 solutions into it, or using the HW1- HW3 solutions we provide. Please refer to the folder structure below as a reference. Remember to source dependencies_ws and finalhomework_ws in every terminal. If you see mushr_sim not found error, source dependencies_ws.

To organize your workspace,

cd ~/Downloads
mkdir -p ~/finalhomework_ws/src/
unzip final_projects.zip
mv final_projects ~/finalhomework_ws/src/

# do this step to copy over code from previous homeworks. 
cp -r homework_ws/src/* finalhomework_ws/src/
cd finalhomework_ws
catkin build

Below is an example folder structure for your workspace setup:

dependencies_ws/
interbotix_ws/
finalhomework_ws/
├── devel/ # You will see this folder after building finalhomework_ws with catkin build.
├── build/ # You will see this folder after building finalhomework_ws with catkin build.
└── src/
    ├── hw1_introduction/
    ├── hw2_kinematics/
    ├── hw3_state_estimation/
    ├── hw4_planning/
    ├── hw5_control/
    ├── final_projects/
    ├── rviz_camera_stream/
    ├── cs4750/
    └──  .../

If you need to re-build and re-source your finalhomework_ws:

cd ~/finalhomework_ws 
catkin clean -y
catkin build 
source devel/setup.bash

Note: Some hints are provided in Hints_for_FinalProject.pdf. You may refer to them as needed.

# Overview

Congratulations on completing previous assignments! The purpose of this "project" is to give you an idea about how previous parts you have implemented come together. In this homework, we have two projects for you to complete. Each of them contains the component of state estimation, planning, and control. We want to demonstrate how these components work together in arms and cars. In general, the planner takes a goal pose or configuration from state estimation methods and then generates a path that will be used by the controller later. The controller executes the path. We will go into the project details and show you how this pipeline works for both the robot arm and the car.

# Arm Final Project (30 points)

It is the year 2050, and robots are widely deployed in dangerous scenarios to replace human workers. In the airport, there is a table conveyer belt, a table, and a trash bin. The robot should be programmed to pick up the items including a tray, a can, and a wallet on the conveyer belt and put them in certain places. First, it should pick up the tray and drop it on the table with a known target pose for pick-up and drop. Then, it should detect the can and wallet with particle filter before pick-up and drop them in the tray on the table and in the trash bin respectively with a known target pose.

In this system, the particle filter or the predefined target pose gives the target for the robot to go to. The planner then computes a path from start to goal. Since the target pose is in Cartesian space, while the planner works in joint space, you need to use inverse kinematics to do the conversion. After this, you need to send the path to the controller, which executes the trajectory.

Task

To complete this, you need to:

  1. Call compute_ik() function to convert from Cartesian space to joint space. See arm_final_project/scripts/airport, in function plan_and_execute.
  2. Obtain the detected value from the particle filter for position of the can and wallet for picking up. See arm_final_project/scripts/airport, under if __name__ == "__main__":.
  3. Insert start and goal joint configuration to the roadmap of the planner. See arm_final_project/scripts/airport, in function plan_and_execute.
  4. Execute the path with controller. See arm_final_project/scripts/airport, in function plan_and_execute.
  5. Under if __name__ == "__main__":, call plan_and_execute to perform the following tasks. See arm_final_project/scripts/airport, under if __name__ == "__main__"::
    • Pick up the white tray
    • Drop it off on the table
    • Pick up the can when the detected position is within a given threshold
    • Throw the can into the trash
    • Pick up the wallet when the detected position is within a given threshold
    • Drop the wallet on the tray

Execute Your Code

Run your code with

# In terminal 1
roslaunch arm_final_project start.launch

# In terminal 2
rosrun arm_final_project airport

Note: Ignore the message [ERROR] [1763522788.350275578]: Semantic description is not specified for the same robot as the URDF

Submit Your Code

For each task, please take a screenshot of your robot right at the time when it is executing the actions of pick-up and drop, and put them in writeup folder. You can see the robot pausing for a short period of time when doing pick-up and drop, during which you can take the screenshot. Name the files accordingly, e.g., pickup_tray.png. We will run the code on our end to ensure the robot works.

Rubric

  • pickup_tray, drop_tray 10
  • pickup_can, drop_can 10
  • pickup_wallet, drop_wallet 10

# Car Final Project: Rescue Cathy (70 points)

It is the year 2050, and autonomous ground vehicles (AGVs) are now used in hospitals. Cathy is a dying patient present in the emergency room. All the doctors are busy, so no one can come to rescue Cathy within the next 200 seconds. Tragically, she will die without medicine in 100 seconds. You are driving the AGV to get the medicine and bring it to Cathy.

The AGV uses a particle filter for car_state_estimation, and there's a planner to plan the path. There are 3 rooms in the hospital, i.e. room A, room B, and room C. In room A, there is a medicine that can extend her lifetime by 50 seconds. In room B, there is a button to call the doctor and let them come earlier by 20 seconds, which means they will come in 180 seconds. In room C, there is a medicine that can add 20 seconds to the patient's life. The AGV can get medicine if it reaches the center of the rooms.

Cathy is considered alive if her lifetime is greater than the time for doctor to come, and the AGV goes back to Cathy before the doctor comes. The AGV can only go back to Cathy once and can not move again after going back to her. Note: The provided example path JSON (final_projects/car_final_project/scripts/sequence.json) is invalid because it visits Cathy twice.

Task

  1. Launch the particle filter, planner, and controller in a launch file. In launch/start_bot.launch, launch the particle_filter node in car_state_estimation package from hw3_state_estimation/car_state_estimation, launch the planner node from car_final_project, launch the controller node from hw5_control/car_controller.
  2. Send the path from the planner to the controller. See src/car_final_project/planner_ros.py, in function send_path.
  3. Implement the is_goal_reached in /src/car_final_project/security_bot.py to check if the AGV reaches the goal pose.
  4. Determine the sequence of room for you to grab medicine in scripts/sequence.json. Each room is represented by a tuple with three elements, [x, y, heading_angle]. Feel free to play with the heading angles, or add waypoints in the middle for a better path. The [x,y] positions for the rooms and Cathy are:
  • C: [14.4, 3.69]
  • B: [1.8, 5.4]
  • A: [3.19, 1.1]
  • Cathy: [9.7, 1.2]

Remember you need to go back to Cathy for these medicines to take effect. You are not allowed to enter any room more than two times, and you are not allowed to enter the same room right after entering it. That means a path like ['A', 'A'] will not be valid.

The json file might look like:

{
    "sequence":[
        [14.4, 3.69, 1.57],
        [1.8, 5.4, 1.57],
        [3.19, 1.1, 0.0],
        [9.7, 1.2, 1.57]
    ]
}

Execute Your Code

After you're done, you should be able to perform the surveillance by running

roslaunch car_final_project start_bot.launch map:='$(find cs4750)/maps/problem-setup.yaml' initial_x:=9.7 initial_y:=1.2 initial_theta:=1.57

Additionally, your implementation of is_goal_reached should pass the tests in python test/security_bot.py.

Try to extend Cathy's lifetime as much as possible. Feel free to play around with any of the components involved. Do not add any additional dependency packages/libraries. Ensure that you are able to run the start_bot.launch with the changes you made.

After the surveillance is complete, you should see an output.txt file under submission. In this file, there is a result of whether Cathy is dead or able to wait until the doctor comes. There will also be Cathy's lifetime, time taken, and the efficiency for extending Cathy's lifetime in this file. The lifetime and time are in seconds. We will create a leaderboard and grade it based on the efficiency for extending Cathy's life, which is lifetime/time. You can only join the leader board if Cathy is alive at the end of the day.

Trouble Shooting

  1. If you see the AttributeError: 'NoneType' object has no attribute 'astype' error, try closing the rviz window and run the program again, or adjusting the waypoints by inserting some waypoints in the sequence.
  2. If you see mushr_sim not found error, source dependencies_ws.

Submit Your Code

Make sure you use your Student ID (SID) as the leaderboard name when submitting. In case Gradescope has any runtime issues, please make sure your code runs correctly in their VMs before you submit and we may verify your solution on your own VM.

Rubric

10 points - Graded based on the leaderboard

30 points - Cathy is alive

20 points - Output file is correct

10 points - Pass all tests in test/security_bot.py