Researchers led by Mohammad Nomaan Husain have built DGF-Asym, a way for a group of robots to agree on an estimate when some sensors are noisy or faulty, according to a paper on arXiv. It extends the DeGroot-Friedkin model of how opinions form, and each robot weighs its peers by how well their past information predicted what came next.
The robots need no prior knowledge of the sensor noise or the kinds of faults, and they do not have to exchange messages at the same moment.
In simulated group navigation with noisy readings and sensor faults that change over time, DGF-Asym gave steadier estimates than three other methods and met the decision rule more reliably. Its decisions took longer, and the results depended on the share of faulty robots, the size of the group and the chosen influence threshold.



