Experimental implementation of robotic sequential nuclear search

In this paper we build on, and extend our previous work on automated nuclear search, by taking the principle of a bidirectional interaction between perception and the sensor management to the implementation level. In the algorithm that we implement in hardware, a model of the underlying physical process is used to guide measurement, which then closes the loop by updating the model. Realization of our "model-driven-measurement" concept leads to a working robotic prototype, capable of confirming the presence of weak radiation sources at the specified statistical confidence level, within a bounded, two-dimensional area. The contribution of this paper is in demonstrating that automated nuclear search is feasible, can be performed with commercially available robotic hardware and measuring devices, and that experimental data are consistent with existing theoretical and computational results.