The objective of this lab is to design a set of controllers to make a robot drive through a set of way points, given to you in a text file, in the presence of unknown obstacles. Figure 1 shows a cartoon of the path that the robot will follow. The blue box is in a known stationary position within the environment, while the purple object will be added by an instructor during the demo. The robot will use onboard odometry and dead reckoning to determine its global position during the navigation. It will be assumed that the robot starts at global position (0m, 0m) with orientation aligned with the x-axis.
Figure 1: Cartoon of the experiment setup. The blue box is fixed in place while the purple faded box will be placed by an instructor during the demo.
For this lab you can develop and test your code running roscore on your computer. However, you must run roscore and all files onboard the robot for the demonstration. To move folders from your computer to the robot, use the scp (secure copy) command. For example,
scp -r <Directory on your cpu to copy>
burger@<ip-of-burger>:<Directory to copy to on robot>
We strongly encourage you to use all available resources to complete this assignment. This includes looking at the sample code provided with the robot, borrowing pieces of code from online tutorials, and talking to classmates. You may discuss solutions and problem solve with others in the class, but this remains a team assignment and each team must submit their own solution. Multiple teams can not jointly write the same program and each submit a copy, this will not be considered a valid submission.
Create a package called TeamName_navigate_to_goal (refer back to the ROS tutorials from Lab 1 if needed). Useful dependencies include rospy, roscpp, sensor_msgs, std_msgs, nav_msgs, and geometry_msgs. You can add as many nodes as you like. An example structure would be:
getObjectRange: This node should detect the ranges and orientation of obstacles. It should subscribe to the scan node and publish the vector pointing from the robot to the nearest point on the object.
Note: You will have to do some filtering of the LIDAR data to determine what measurements of the 360 are useful. You may also want to segment your readings to be able to discern two obstacles apart from one another. You will only encounter one obstacle at a time, but if your LIDAR sees the wall or a stray chair/other robot you will want it’s object estimate to be robust. It is also important to remember that this data is with respect to the robot’s local coordinate frame.
goToGoal: This node should subscribe to the odom node which determines the robots global position from onboad sensors for you (using dead reckoning). It should also subscribe to the getObjectRange node to determine if there are any obstacles that need to be avoided.
This node should first read in the given goal locations from the wayPoints.txt file, or you can include them in your code some other way. You should then create several controllers that drive the robot through the sequence of given goal points without colliding with unknown obstacles. To receive full credit the robot must stop for 10 seconds within a 10cm radius of the first goal point, 15cm radius of the second goal point, and 20cm radius of the third goal point, the robot must not hit any obstacles, and the robot must reach the destination in under 2 minutes 30 seconds.
-
Remember the onboard odometry and goal points are given in the same global frame while the measurements are in the robot’s local frame. The 2package tf2 in ROS (http://wiki.ros.org/tf2) may be useful to transform coordinate frames if you want but is not necessary for this lab.
-
The Turtlebot3 has built in odometry which you are free to use. You can access it by subscribing to the /odom topic. It relies on proper calibration beforehand which can mess up if you move the robot during its bringup. It is highly recommended to print out the robot’s estimated pose to make sure the odometry is correct and not drifting while the robot is stationary. If you find it is messed up it can be fixed by placing the robot on the floor and restarting the bringup.
-
The odometry node saves the current position of the robot and starts where it left off. If you pick up the robot and restart your program to run the course, the odometry given to the robot will be the position and orientation the robot was last at before you picked it up. We have given you a python script (RotationScript.py) which records the initial odometry readings and subtracts them as an offset so your assumed starting position is the origin with heading aligned with the x-axis. You may integrate this into your project however you want.
-
The angular component of the odometry is represented by a quaternion which should be used appropriately.
-
If you wish to create dead reckoning position updates yourself, or augment the ones produced in the /odom topic, you can access the IMU and encoders through published topics /imu and /sensor state. More details can be found at http://wiki.ros.org/turtlebot3_bringup#Published Topics.
You are allowed 5 attempts to demo this to an instructor and will receive the best score of your attempts.
| Criteria | Grades |
|---|---|
| Run the code onboard the robot | 25% |
| Drive within 10cm of the first goal point | 25%( e-stopped_distance_outside_of_goal_in_cm ) |
| Drive within 15cm of the second goal point | 25%( e-stopped_distance_outside_of_goal_in_cm ) |
| Drive within 20cm of the third goal point | 25%( e-stopped_distance_outside_of_goal_in_cm ) |
| Each collision with an obstacle | -5% |
| Take more than 2 minutes, 30 seconds to reach the final goal point | -15% |
You run all your code on the robot. Your robot reaches the first goal point within 10cm, hits the obstacles once but makes it within 20 cm of the secondgoal point, and then reaches the final goal point within 20cm. This is all done within 2 minutes, 30 seconds. Your grade would be...
grade = 25 − 5 + 25e-5/25 − 25 + 25 + 25
= 25 − 5 + 20.5 + 25 + 25
= 90.5
You have two required submissions for this lab.
1 Your ROS package in a single zip file called Lab4_LastName1_LastName2.zip uploaded on Canvas under Assignments–Lab4.
2 A live demonstration of your code to one of the instructors. This can be done anytime before the due date at the top of this lab. Class will meet in the lab room on the due date to allow everyone to demo their controllers. If you demo before the due date you do not need to come attend class that day.