Skip to content

testing_and_troubleshooting.md

Latest commit

 

History

History
56 lines (34 loc) · 2.52 KB

testing_and_troubleshooting.md

File metadata and controls

56 lines (34 loc) · 2.52 KB

Testing & Troubleshooting

🛠️ Common Issues & Fixes

Below are some known issues you might encounter when working with ARI, along with suggested troubleshooting steps.

Robot Not Moving After Goal Sent

If a goal has been sent to move_base but ARI isn’t moving, it’s likely due to a crash or malfunction.
Fix: Check the Web Commander interface at http://<ARI_IP>:8080 to inspect running processes and ensure the cameras and other critical modules are functioning properly.

Multibody System Not Responding

If the multibody system isn't functioning when ROI, ID, and cropped image messages are being sent:
Fix: Verify that these messages are published with correct real-time timestamps. Proper synchronization is essential for the system to work.

Face Recognition Errors (YOLO)

Face recognition using YOLO is not foolproof—blurry frames can lead to incorrect person entries, causing issues over time.
Fix:

  1. SSH into the robot: ssh pal@ari
  2. Navigate to the folder: deployed_ws/lib/dd2414_human_detection/
  3. Delete the existing JSON file.
    A new file will be generated on the next face recognition cycle, but note that this will reset all previously stored person data.

Head stops moving

When a frame is published to the look_at topic which the robot can’t turn the head to, expressive_eyes crashes. This can happen when ARI is moving while turning the head.

Fix:

  • Check the Web Commander interface at http://<ARI_IP>:8080 for expressive_eyes. If it has stopped running, restart it.

Navigation in Narrow Spaces

To navigate tight areas like office doors, ARI’s obstacle radius is dynamically reduced in designated "door zones."
Note:

  • Ensure all narrow doors are marked as door zones in the map (“door_XX”).
  • This allows ARI to turn fully and pass through safely without collisions.

Accessibility Limitations

Most rooms were accessible via ramps; however, the MOCAP room contains a step that ARI cannot cross—even with a ramp.

Microphone Limitations & Speaker Detection

ARI’s microphones are arranged in a circular array at the front. This setup struggles with locating speakers behind the robot, especially in small, echo-prone rooms.
Solution:

  • Speaker detection is handled via multimodal integration.
  • ARI uses the microphones only to detect the presence of speech and estimate the turning direction.
  • She stops turning once a person is visually detected using the head camera.