Testing & Closing Issues
Index
Testing on Windows
First of all we will install docker on windows.
Once installed we will download the image we are going to test with.
1
docker pull jderobot/robotics-backend:latest
To use the docker command in the windows terminal, we have to open the terminal as an administrator.
Once the docker image is downloaded, we can run it as the example in unibotics.
When we launch it, we can test unibotics or robotics academy locally. After several tests from windows I have verified that all the exercises work perfectly, but they seem to run slower than from linux. I have also noticed that the evaluate code
functionality of unibotics is not working.
GPU Acceleration
SO: Windows 11, RAM: 16 GB, CPU: AMD Ryzen 5 7535HS (6 cores), GPU: NVIDIA GeForce RTX 3050 Ti:
(Average Percentages up to 100%)
Task | Gazebo’s RTF | Gazebo’s FPS | % CPU Usage | % GPU Usage |
---|---|---|---|---|
Follow Line | 0.74 | 23 | 82% | 6% |
Vacuum Cleaner | 0.98 | 19 | 61% | 6% |
AutoParking | 0.84 | 9 | 70% | 7% |
Follow Person | 0.97 | 25 | 70% | 5% |
Localized Vacuum Cleaner | 0.98 | 20 | 71% | 6% |
Global Navigation | 0.85 | 23 | 72% | 9% |
Rescue People | 1.00 | 34 | 77% | 11% |
Obstacle Avoidance | 0.80 | 16 | 74% | 5% |
3D Reconstruction | 0.90 | 14 | 80% | 7% |
Amazon Warehouse | 0.98 | 7 | 82% | 6% |
Montecarlo Laser Localized | 0.98 | 20 | 75% | 12% |
No GPU Acceleration
SO: Windows 11, RAM: 16 GB, CPU: AMD Ryzen 5 7535HS (6 cores):
(Average Percentages up to 100%)
Task | Gazebo’s RTF | Gazebo’s FPS | % CPU Usage |
---|---|---|---|
Follow Line | 0.54 | 7 | 80% |
Vacuum Cleaner | 0.98 | 11 | 75% |
AutoParking | 0.75 | 4 | 78% |
Follow Person | 0.95 | 4 | 79% |
Localized Vacuum Cleaner | 0.97 | 11 | 74% |
Global Navigation | 0.83 | 15 | 75% |
Rescue People | 0.99 | 14 | 78% |
Obstacle Avoidance | 0.74 | 10 | 76% |
3D Reconstruction | 0.86 | 6 | 79% |
Amazon Warehouse | 0.96 | 4 | 79% |
Montecarlo Laser Localized | 0.97 | 12 | 74% |
Testing issue #2667
After several tests, it seems that the port error has been fixed.
Testing Digit Classifier
To test this exercise implemented by a colleague, create a miniradi with the following command:
1
sudo ./build.sh -m issue-137 -a digit_class_migrate -i issue-411 -t mihit-test-dig
Also editing compose_cfg.
After creating it, we launch rb and test it. For some reason Ubuntu 22.04 has an open bug with my webcam, I had to download a kernel patch to fix it repo.
After trying the exercise, I get the same error as my colleagues.
issue #2464
There are several people working on this issue looking for why a widget is not displayed correctly. After some testing I discovered the source of the problem and even managed to recover the time stamp functionality.
After checking other residual files in the repository, I managed to get everything working, but for some reason it didn’t work 100% of the time. Here is my progress: draft.
A colleague managed to solve the problem occurring to my progress by another route and the issue has been resolved.
issue #2695
Updating legacy ports, now the developers enviroment uses tonly 4 ports. (Clossed)