Crystal is an interference-resilient ultra-low power data collection protocol for wireless sensor networks (WSN) especially efficient in applications generating sparse aperiodic traffic. It was shown to achieve over 99.999% packet delivery ratio in presence of Wi-Fi interference with per-mille radio duty cycle in some real-world applications. Crystal uses Glossy as underlying communication and time-synchronization primitive.
Crystal got the 2nd prize at the EWSN'18 Dependability Competition!
- Data Prediction + Synchronous Transmissions = Ultra-low Power Wireless Sensor Networks, Timofei Istomin, Amy L. Murphy, Gian Pietro Picco, Usman Raza. In Proceedings of the 14th ACM Conference on Embedded Networked Sensor Systems (SenSys 2016), Stanford (CA, USA), November 2016, PDF
- Interference-Resilient Ultra-Low Power Aperiodic Data Collection, Timofei Istomin, Matteo Trobinger, Amy L. Murphy, Gian Pietro Picco. In Proceedings of the International Conference on Information Processing in Sensor Networks (IPSN 2018), PDF.
Currently the protocol is implemented for the TMote Sky and CC2538 platforms only. Please contact us if you need the code for CC2538.
The master branch is the current stable branch of Crystal, while the devel branch has the most recent modifications.
Please contact us in case you need the exact code we used in IPSN'18 experiments. The version we used for the EWSN'18 Dependability Competition is instead published here for reference (depcomp18 branch). It contains several competition-specific tweaks and will not be updated in the future.
Disclaimer: Although we tested the code extensively, Crystal is a research prototype that likely contains bugs. We take no responsibility for and give no warranties in respect of using the code.
You can find a very simple example in apps/crystal-test-simple. Refer to the README file there for building instructions for Cooja and TMote Sky. Another example, apps/crystal-test is an application similar to the one we used in the SenSys'16 and IPSN'18 studies. It is used by our scripts to generate test jobs for testbeds and process the collected logs (described below).
To run experiments with different sets of parameters you can build a binary (or a set of binaries) using the test_tools/simgen_ta.py script. It reads the parameter set(s) from params.py file.
A good starting point for defining your parameter set can be found in the exps/example/ directory. You may copy the whole directory and edit the parameters and the list of nodes present in the testbed. After that, run ../../test_tools/simgen_ta.py. It will create one or several subdirectories (if they don't exist) named after the individual parameter sets defined in the params.py file and put there the binaries compiled from apps/crystal-test.
You can specify "cooja" in the testbed name to compile for Cooja, list the node IDs, and use MRM radio model for simulations (e.g., one defined in apps/crystal-test/mrm.csc).
The repository has submodules for fixed Cooja and MSPsim that are recommended for Glossy/Crystal. The patch fixes a problem in the CC2420 emulation of MSPsim that prevents Glossy from reading the packet from FIFO while it is being received. This allows running tests with longer packets (up to 75 bytes) in Cooja. With unpatched Cooja only very short packets are supported (few bytes). Note that if you want to use the unpatched Cooja you need to define PATCHED_COOJA=0.
To use Cooja you will need JDK 8. To install and compile Cooja, do the following:
git submodule update --init
cd tools/cooja
git submodule update --init
ant jar