Automatic Parameter Exploration for Low-Power Wireless Protocols

We present APEX, a novel parameter exploration framework for low-power wireless protocols. APEX can autonomously derive an optimized set of parameters allowing a protocol to satisfy certain application requirements on the reliability, efficiency, or latency of communications. It does so without the need of expert knowledge, and by minimizing the number of testbed trials executed to gauge the protocol’s performance as a function of different parameter combinations. We have created a preliminary implementation of the framework coupled with the D-Cube testbed, and used it to parametrize the Baloo-Crystal protocol for different application requirements. Our results show that APEX can find an optimal parameter set with as little as 13 testbed trials in the median case, and reduce the average experimentation time by 65% compared to approaches based on exhaustive search.