Show HN: Help test out swarm robot control | Hacker News Reader

Show HN: Help test out swarm robot control

swarmcontrol.herokuapp.com

31 pointsangersock12 comments

This experiment is the result of several months of back and forth with my good friend and collaborator Aaron.

The goal is to test several different control and visualization schemes for swarms of robots, and to gather quantitative data about what mechanisms help people work with these swarms most effectively.

Secondarily, we've created a simple platform for publishing these academic UX projects online--why settle merely for a handful of undergrads to sample with? We'd like to advance the state of the art a bit in experimentation.

All results are available as CSV or JSON from the results page, and we'll be posting up a link to the project source soon so that other people can learn from our work and extend it.

Thank you everyone here for giving me the feedback and motivation to start doing web stuff--without this community, I wouldn't have either the skills or motivation to pursue fun projects like this.

EDIT: We appreciate any feedback that we get, good or bad. Thanks again!

Totaly OT, but I feel obligated:

"FOR SCIENCE" You monster. (Portal 2 reference. You should all play this game.)

They say great science is built on the shoulders of giants - not here. At Rice University we do all our science from scratch; no hand holding.

I proposed an approach for human swarm interaction in my Master's thesis [1] and published a conference paper based on the research [2]. I wonder if you have come across my work and would like to discuss it if you are interested.

[1] https://mospace.umsystem.edu/xmlui/handle/10355/27879 [2] http://faculty.cs.byu.edu/~mike/mikeg/papers/MOSC/Bashyal_09...

Ha, cool work! We're always happy to chat about our research--just hit us up on the linked support email or through my HN profile.

We're trying to get something together in time for ICRA, but we had so much fun fiddling with the sim we almost blew our deadlines. :)

Your paper is interesting. Can you share a link to a video of your simulator in action? Better, is the code still live? It appears that you allowed the user to switch modes between global commands and "attract/repulse". What modes did the users prefer?

The term "assistive swarming" is intriguing. In future work, we'd like to simulate interacting with an intelligent swarm, perhaps a group of Boids (http://en.wikipedia.org/wiki/Boids). Our current work is inspired by micro/nano setups by colleagues here at Rice, and RPI, Drexel, and Harvard -- see the videos page for an overview: http://swarmcontrol.herokuapp.com/videos

Here is the video for radiation source localization problem: http://www.youtube.com/watch?v=HKofA3ygvPA

And here are two other use cases: http://www.youtube.com/watch?v=49rxqItJFeY http://www.youtube.com/watch?v=jrq6nq3FLmo

This was all done in Matlab as that was the preferred platform for my lab.

Nice -- the transparent paths are effective for showing the explored area.

Lately I've started sharing code from my papers -- is your code available? Matlab code is easily shared at MatlabCentral: http://www.mathworks.com/matlabcentral/fileexchange/authors/... (my page).

What research result surprised you the most when doing your user experiments? Can you suggest an experiment you'd like to see at http://swarmcontrol.herokuapp.com ?

Are the input constraints (e.g. all robots move the same way at once with the same force) entirely in the presentation of interface to the user, or are there technical constraints that mean the robots all have to behave the same way?

Water can flow, or it can crash. To push the polygon to the right you want to flow past it to the left, then crash against it to the right. If there were a 'move without pushing' it would make the flowing stage a whole lot simpler.

So, for real-world applications, we have to deal with some constraints, right?

An obvious physical analogue would be iron-filings suspended in a fluid, with magnetic fields "commanding" all of them to move left, right, etc. In that case, we obviously can't just have the filings move through each other or the workpiece being manipulated.

Two extensions to our control schemes are having alternating implode/explode movement (move to center of flock, move away), and also exert a torque (which is simple to do; if every robot in the swarm has an orientation and knows whether the swarm center is to its left or right, we can have them move forward or backward, and then we get torsion).

It's a really exciting field--email us!

To answer another part of your question "are there technical constraints..", our research has mostly been inspired by practical challenges in micro and nano robotics. At such small scales, it is difficult (presently) to do onboard computation and these robots are generally controlled by a global field that is electric/magnetic/photonic/chemical gradient/gravitational etc.

The 'move without pushing' might require a state change (temperature dependent viscosity?). Perhaps we could use 3D robots that have shorter length scale in some direction than the others, and flip them sideways for flowing. Any suggestions?

Reply on news.ycombinator.com