Student Exercises

This is a set of student lab work exercises which are used in different courses. As these lab exercises are embedded in courses, they have a different focus and a different order than the tutorials. The student lab work exercises supplement material and refer to the tutorials whereever reasonable.

Typically, the software infrastructure is already setup in your lab. Talk to the lab staff and you get the credentials in the lab. Go ahead with the exercises.

In case you need to setup the software infrastructure on your own or in case you want to run the software on your own computer, please follow the instructions . There you find the instructions for downloading a ready-to-run virtual machine and also the link for a direct installation onto your PC without a virtual machine.

We use a simple laser-based obstacle avoidance algorithm with a simulator to make the first steps with the SmartMDSD toolchain. You then implement a laser-based hallway following algorithm, a laser-based wall-following algorithm, and finally a laser-based algorithm for detecting and approaching a cylinder.

Your task is to figure out how the laser obstacle avoid algorithm works. Can you explain what is does?

You learn:

• how to start a ready-to-run system
• how a simple laser-based obstacle avoidance algorithm looks like
• how the internals of a component look like
• how to access the ports of a component from inside a component
• how to recompile a component in case you did modifications

Follow the tutorial Simple System: Laser Obstacle Avoid and use the following versions:

The following gives some more information on the laser data communication object:

• look also at the documentation inside | CommMobileLaserScan.hh
• a laser scan is a collection of rays (or scan points i.e points in the laser scan)
• the resolution of a scan is the angle between consecutive rays
• a ray with zero angle aligns with the robot heading
• each scan point has a distance, an index, a reflectance (not used in simulators)
• the start angle, resolution, and index are used to determine the angle of the ray: angle of the ray = (start angle) + (resolution) * index

Your task is to let the robot move in the middle of the hallway. For this, use the laser ranger, calculate motion commands and command them to the robot. We provide two different worlds which come with different challenges for your algorithm.

• You get more familiar with the characteristics of a 2D laser ranger
• You gain more experience in modifying, compiling, deploying and running a system with the model-driven toolchain

Your task is to let the robot follow the left / right wall. For this, use the laser ranger, calculate motion commands and command them to the robot. Again, we provide two different worlds which come with different challenges for your algorithm.

Your task is to let the robot approach a cylinder (move to a detected cylinder and stop in front of it). For this, use the laser ranger, calculate motion commands and command them to the robot.

• the diameter of each cylinder is 15cm
• there is a depth jump in the laser scan points at the left edge and the right edge of the cylinder. This can be used to detect a cylinder
• be aware, your algorithm should work still in case there are several cylinders in the scenario

We now use an Intel Realsense RGBD sensor instead of a laser ranger. A RGBD sensor provides a colour camera image (RGB) and a depth image (D).

This exercise shows how you can use a depth camera (in our example, an Intel Realsense RGBD camera) instead of a laser ranger. We illustrate the use of the depth camera by the example of the obstacle avoidance algorithm which so far used a laser ranger. We added another component which converts the 2-dimensional depth image from the RGBD camera into a laser scan.

We use the following fully preconfigured system: SystemExerciseRGBDObstacleAvoid.

After starting this system, the robot moves around and avoids obstacles based on the depth image of the RGBD camera. The world used in this scenario is shown in the figure below.

Please go to the System Builder perspective and see which components are used now and how they are wired:

• ComponentWebots3DCamera generates color and depth images (CommRGBDImage) and sends them to ComponentLaserFromRGBDServer
• ComponentLaserFromRGBDServer converts the depth image into the laser data format (CommMobileLaserScan) and sends it to ComponentLaserObstacleAvoid. Basically, the conversion is a down-projection of the depth image to result in a planar laser data format (all the scan points are in a horizontal plane)
• ComponentLaserObstacleAvoid calculates velocity commands for collision free motions and sends those (CommNavigationVelocity) to ComponentWebotsMobileRobot
• ComponentWebotsMobileRobot drives the mobile robot according to the given velocities

What are the take home messages from this exercise?

• Please notice that there is no connection anymore from ComponentWebots2DLidar to ComponentLaserObstacleAvoid. The laser ranger is not used anymore for obstacle avoidance.
• While the laser ranger sees obstacles only in a horizontal plane, the depth camera has a different field of view which allows to detect more obstacles. For example, in the used world, the laser scanner is mounted too low to detect the table top while the depth camera is able to detect it and to avoid it.
  • Press CTRL + F8 in Webots to show / hide the rays of the laser ranger and to check what part of the world it sees and which not
• Although we are now using a different sensor, we can reuse the laser obstacle avoidance component without any modifications at this component.
  • of course, it just provides a port for receiving laser scans but we can connect any component that has a port for providing laser scans. For this, we just added another component in between the RGBD camera component and the laser obstacle component.

For this exercise, we use the following versions:

Similar to Exercises: Laser Ranger / Lab 4: Approach Cylinder, we now want to approach a cylinder. The difference is that we now use the RGBD camera:

• detect a cylinder of a given color via the colour and / or depth image
• approach the cylinder and stop in front of it

The world used in this exercise is shown below:

Your task is to write the corresponding software:

• add your code in ComponentExerciseRGBDApproachCylinder/smartsoft/src/RobotTask.cc
• the library OpenCV 4'is available to process images

The following labs dive deeper into navigation. We start with ready-to-run system projects that involve a whole bunch of software components related to different aspects of navigation.

«to be added»

This lab gives you insights into a ready-to-run navigation stack. The navigation stack comprises components for path planning, for map building, for motion control, for localization, for task coordination, and for world models. The objective is to first get an overview on navigation capabilities and gain some experience with navigation capabilities before we go into details.

Please follow A More Complex System using the Flexible Navigation Stack.

«to be added»

«to be added»

This tutorial is not part of a series.

Contributions of Nayabrasul Shaik, Thomas Feldmeier.