Hacker News Reader

Comment by jimiasty | Hacker News Reader

Hi Matti,

there are several approaches to Bluetooth Low Energy beacons. Project called Tail (http://www.thetileapp.com/) is building sensors that you attach to products and when they move or are lost your smartphone can use the Bluetooth radio signals to detect its proximity.

Our www.estimote.com beacons are designed to be placed in a fixed location such as retail store, museum or park. They broadcast their location to all the phones or devices (like your robot) in the range.

The devices could pickup the signal and estimate the location very precisely, down to inches - so it is definitely a technology that could be used to indoor-navigation for robots, machines in the factory or doctors in a hospital.

We at Estimote are focused on retail, but we also have an API and presale Dev Kits, so anybody could experiment with the technology.

Shoot me an email jakub at estimote.com and will be more than happy to advice on how to use it for robots.

View on HN

> The devices could pickup the signal and estimate the location very precisely, down to inches - so it is definitely a technology that could be used to indoor-navigation for robots, machines in the factory or doctors in a hospital.

There has been a _lot_ of research in this area that I have seen and almost everything seems to ignore multipath propagation. I think the more interesting approach is fusing RF localization with other sensors (example paper [pdf warning for those fearful of them] https://smartech.gatech.edu/bitstream/handle/1853/37074/perv... )

Just wanting to point out that it is much more complicated than simply picking up the signal and homing in.

You are definitely right and there are technology limitation just to stick 3 sensors and triangulate the signal.

We believe the whole indoor-navigation idea is wrong. There is no need to map every inch of the physical world and get X,Y coordinates.

It's all about the context and placing tiny beacons in the doors, next to check-out, next to GoPro camera in the Best Buy.

They you detect the proximity and the context and trigger an action.

Of course if you would like to navigate down to inches and get X,Y you could just narrow the range of beacons and stick as many as possible - you will be very accurate and the multipath propagation is not an issue.

I'm first author on that Pervasive article... my PhD thesis focused on using passive UHF RFID tags affixed to objects, people, etc to allow mobile robots to quickly take inventory and then approach (locate) tagged objects in the home [1].

Passive UHF RFID tags are nice for a number of reasons: super-low tag cost (sub-$0.10 in bulk), long range (6+ meters), and lack of a battery (tags harvest wireless power from a reader). Bluetooth solutions will have a hard time competing with UHF RFID on these properties -- and that's key for certain applications. However, there are definitely benefits to an alternative solution like wireless triangulation. For example, even the best UHF tag localization algorithms produce error bounds on the order of 0.5 meters (there are workarounds for mobile robots). Plus, the Bluetooth solution doesn't require the reader to be mobile -- you can take a single measurement and get a decent pose estimate.

I still share a lot of your concerns: if this is using triangulation, you'll need multiple base stations (ie. infrastructure costs); many similar systems require extensive calibration; and the effects of multipath, diffusion, and fading can be very tricky to characterize.

As for mapping... SLAM mapping using ultra-low cost laser rangefinders [2] is nearing triviality, and even onboard visual SLAM is becoming imminently feasible. Anyway... happy to speak more offline. My contact info is easy to find.

[1] More details: http://www.travisdeyle.com/publications.html

[2] http://www.hizook.com/blog/2009/12/20/ultra-low-cost-laser-r...

I happened to work with one of your co-authors a handful of years ago. Did you happen to use a ThingMagic reader? I worked on the software side of the fixed readers (arm-linux based), not the small embedded reader modules. I also added a lot of improvements to the python API for rapid prototyping using the embedded module.

Indeed, I used the ThingMagic M5e. All the code for the (python) drivers [1] and robot behaviors [2] are open sourced.

Funny story... before (and after) doing higher-level robot behaviors, I worked on low-level reader hardware too (with our mutual colleague, I presume). Most recently, it was building SDRs to interact with super-high datarate passive tags with onboard sensors... to build cyborg dragonflies [3].

[1] http://ros.org/wiki/hrl_rfid

[2] http://www.ros.org/wiki/pr2_rfid

[3] http://www.wired.com/wiredscience/2013/06/dragonfly-backpack...

I forgot that the stock python drivers weren't open source. The "dragonfly backpack" is a cool looking project.

Reply on news.ycombinator.com