Loona Programming - Obstacle Detection

[philb1010]

Has anybody written a program that uses obstacle detection? How well does it work?

With Turn on visual recognition Obstacle detection near the start and Obstacle detected within 30 cm ahead inside the main loop, it only sometimes works (she backs up as instructed). Other times she runs into a wall or something (light or dark color, doesn't seem to matter) and tries to keep going. I'm pretty sure the logic is correct.

What's been your experience?

 

[Chris]

Has anybody written a program that uses obstacle detection? How well does it work?

With Turn on visual recognition Obstacle detection near the start and Obstacle detected within 30 cm ahead inside the main loop, it only sometimes works (she backs up as instructed). Other times she runs into a wall or something (light or dark color, doesn't seem to matter) and tries to keep going. I'm pretty sure the logic is correct.

What's been your experience?

If you can make a program available for download in the Hello Loona app and post a screenshot of the program here I can run it on each of my Loona and report back to you if I’m getting the same or different results.

What do you think the problem is? Hardware related? Very interesting observations and hope to hear more about why this might be happening.

 

[philb1010]

I have the Loona app for Android and see no option to share programs, but I am including a screenshot here, a simple program to demonstrate obstacle detection. Most of the time it works and she backs up and changes direction, but once in a while she doesn't see a wall or object and just keeps trying to go through it. Run the program long enough and she'll probably get stuck against a wall or object. I have been doing experiments like this indoors in moderate to bright lighting conditions with varying objects and surroundings, Loona's head angle, etc., with the same results.

Here's another experiment you can try with this program: Have her facing a wall or object, perhaps a dark-colored one, either touching it or very close, and start the program (on Android tap Execute). She may not detect the obstacle and just try to move forward, forever. (For me this situation is easy to reproduce.) Now stop the program (on Android tap Operate to stop it), and notice her default/built-in program/behavior will immediately back up because she see's the obstacle. This shows that she can in fact see obstacles better when she's running her default program/personality versus what is available for object detection in user programming. I can leave her running on her own like this for a very long time and she never gets stuck anywhere.

I wonder if there is a bug where maybe the 3D ToF LiDAR Sensor does not work when running a user's custom program and only works when she is running her default, built-in program. Perhaps only the camera is used -- and not LiDAR -- for obstacle detection when running a user program?

Anyway, please let us know your findings. Thanks a lot.

Here's another link in case image doesn't appear below: https://photos.app.goo.gl/3wyg8G3wdEqav12u8

 

[Chris]

I have the Loona app for Android and see no option to share programs, but I am including a screenshot here, a simple program to demonstrate obstacle detection. Most of the time it works and she backs up and changes direction, but once in a while she doesn't see a wall or object and just keeps trying to go through it. Run the program long enough and she'll probably get stuck against a wall or object. I have been doing experiments like this indoors in moderate to bright lighting conditions with varying objects and surroundings, Loona's head angle, etc., with the same results.

Here's another experiment you can try with this program: Have her facing a wall or object, perhaps a dark-colored one, either touching it or very close, and start the program (on Android tap Execute). She may not detect the obstacle and just try to move forward, forever. (For me this situation is easy to reproduce.) Now stop the program (on Android tap Operate to stop it), and notice her default/built-in program/behavior will immediately back up because she see's the obstacle. This shows that she can in fact see obstacles better when she's running her default program/personality versus what is available for object detection in user programming. I can leave her running on her own like this for a very long time and she never gets stuck anywhere.

I wonder if there is a bug where maybe the 3D ToF LiDAR Sensor does not work when running a user's custom program and only works when she is running her default, built-in program. Perhaps only the camera is used -- and not LiDAR -- for obstacle detection when running a user program?

Anyway, please let us know your findings. Thanks a lot.

Here's another link in case image doesn't appear below: https://photos.app.goo.gl/3wyg8G3wdEqav12u8

Incredibly interesting observation about the LiDAR not working using a custom program.

I tried running a very simple program and found Loona responded to my hand in front of her but also found she failed to respond sometimes. This was in low light. You have greater patience than I, for me Loona was frustrating me with the loud noises and wheel banging trying to test this out late at night and I gave up.

Yes I have noticed Loona will try to move forward against dark objects indefinitely, I have a video on YouTube where one of mine is trying to go forward at full speed into a dark space underneath a lounge chair, and is literally burning rubber, I find it quite hilarious.

You could try asking people in the Loona community in the Hello Loona app this question and get them to run your program and probably get a better response and richer discussion than what I can generate as a single test case. Sorry I can’t be of much help-I’m too impatient to do programming.

To upload a program to the community follow these steps:

Go to Community > Publish > Upload Program



You can then upload media content such as a video of Loona alongside your custom program and pose your questions. I’m sure you will generate a lot of discussion. Thanks for stopping by and sharing your observations.