Tracking of Dual Gas Concentration Using Carbon Nanotubes Based Reservoir Computing

Reservoir computing (RC) has the advantages of fast learning and low training cost, as only the readout weights need training. Furthermore, it is suitable for multi-task processing or continuous learning as there could be no interferences between tasks. In this paper, we prepared a carbon nanotubes (CNTs)/PBMA reservoir modified with phosphomolybdic acid molecules (POM). The dynamical tracking of dual gas concentration in the environment was carried out by our lab-made RC experimental platform with the CNT reservoir under atmosphere. Through grid search of several RC parameters in this implementation, we obtained the optimized minimum tracking accuracy of less than 1 ppm for NH
3
and NO
2
when they were randomly mixed under a closed measurement environment or an open environment. This demonstrates a very promising potential of the POM/CNT/PBMA reservoir to perform near/in gas sensor computing.