Stringman Installation Vid and Progress Update

[Nathaniel Nifong]

Stringman reels closer every day to being a "turn it on and forget" clutter pickup tool.

The two latest models naavox/targeting and naavox/visual_servo are responsible for finding and grasping targets respectively. These are actually very small models compared to what I had been working with before, because I have found that predicting specific meaningful variables from training data such as how much pressure to eventually grasp with, is more effective than letting the model imitate how I drive.

You can see an example of these here though the attempt didn't clean up much (yet)

Installation Video

If you've purchased Stringman and need a visual guide on how to install it, there is now a video covering the whole process.

The merged floor camera view now has an improved blended appearance similar to that used by the top-down surround view featured in boats and cars.

It's also now possible to share a robot with "everyone" and when you do this, literally any logged in user can drive it, with a max of 3 at a time. anyone after that gets a spot in a queue and a scalable HSL stream for spectating.

Instead of just honking, anchors connecting to a new wifi network now intone the last three digits of their IP address.

More demo videos coming soon! let me know what you'd like to see!

As always, I am eagerly looking for new locations where I can bring a Stringman robot to demo and collect diverse training data. If you'd like to try it out, let me know.