Well, Hmm. “Project Tango crashed”

I’ve had to reboot maybe 8 or so times after gtting the device prepped according to Google. The hardware is interesting and very cool. Is it ready for prime time? I don’t know yet. This screen is my most familiar:


And then it crashes. Yuck. But…. After rebooting, and SUCCESS.

IMG_20141211_161832 IMG_20141211_161836 IMG_20141211_161744 IMG_20141211_161751 IMG_20141211_161757 IMG_20141211_161800

This is the program Project Tango Explorer. It has 3 modes: Area Learning, Pointcloud, and full diagnostics. The area leaning is what seems to crash consistently on me.  I tried to map areas and it usually fails on about 2-5 minutes with Tango crashed. I moved on to pointclouds.. and that’s what you see up there. The thing has an infrared laser on it.. And weird! I’ll be submitting a video shortly through Youtube. It looks like it’s a Time of Flight laser, given the interferometry pulses.


5 responses to “Well, Hmm. “Project Tango crashed””

  1. Alexander Grau says :

    I can confirm that Tango Core crashes even in Lenovo PHAB 2 Pro. It lives longer already (20 minutes), but then it crashes.

    • jwcrawley says :

      Egads.. I don’t know what they’re doing that’s causing the incessant crashes. But I do know that this platform is completely worthless.

      I’ve thought about something like a USB-OTG with a Kinect running LSD-SLAM with ROS. My experiences with ROS and LSD-SLAM just don’t crash like that…. It may lose telemetry and mapping, but the underlying nodes just don’t go “SEGFAULT”.

      And ideally, I’d like to see more robustness with ROS, primarily pertaining to roscore crashes and redirection of roscore to failover machines. That doesn’t exist currently, but is an essential feature. Think of it like this: a robot needs to know locality and details of other bots. So when it comes within area of a new roscore, it can interrogate it for local details.

      As an aside, one of my current projects is to convert a 3d printer to ROS… Then give it vision and a whole bunch more cheap sensors and give it a form of intelligence. I’m surprised no one is doing this yet.

      • Alexander Grau says :

        My ‘problem’ is that nothing currently has such a localization robustness as Tango (Here you can see a working prototype, the final scenario is outdoor: https://www.youtube.com/watch?v=w7HX3Q3A47w). I have tested rovio and LSD-SLAM, they need a lot more work to get close to Tango’s performance outdoor. But with some tweaking it should be possible. Also, I think I found the reason at least for my own App crashes: in the renderer, I’m creating objects too quickly, at least more quickly the garbage collection can free. This will generate a crash after 20 minutes. So, hopefully this was the only problem 🙂

      • Alexander Grau says :

        Hmm…now you made me curious – why in the hell would someone convert a 3d printer to ROS? 🙂 And btw, I found out rovio is working very similar to Tango’s tracking: http://grauonline.de/wordpress/?page_id=2014

  2. jwcrawley says :

    Well.. You’re indeed right that LSD SLAM isn’t robust. That’s because LSD-SLAM only does gradient descent to determine position and travel. When you add another source of data, like a 3/9DoF accelerometer, you can greatly improve your tracking to that of Tango.

    Now, as for making a 3d printer use ROS: 3d printers have limited sensor systems. And because of constrained sensors, many things can go wrong with a print job. But if one were to have lots of sensors, we could make decisions about of the printjob is going right or wrong. Not only that, we could also incorporate a visions system to overlay an STL over the camera input. That would allow us to determine % accuracy, along with detection of other issues (shifted layers, air printing, spaghetti,….).

    But yeah, just a few ideas and things I’m working on.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: