The confront of medical ethics, 'rational allocation of scarce medical resources' practically appears as uncertainty in secondary-spectrum-usage in a cognitive Body Area Network (BAN). In this context, we have formulated RMSRA-VB (Rational Medical Scarce Resource Allocator Virtual Backbone) as an initial step to provide autonomy in wireless agents so that they can adapt to that uncertainty caused by collision-avoidance and EMI (Electro-Magnetic Interference)-avoidance. Then, we have proposed a distributed autonomic learning framework, RMSRA-QL-POSG(RMSRA using Q-Learning in a Partial Observable Stochastic Game Model). RMSRAQL-POSG is the first Q-learing algorithm in a POSG Game Model. By RMSRA-QL-POSG, wireless agents learn to utlize secondary-channel in uncertain cognitive BAN. Our probabilistic analysis proves how RMSRA-QL-POSG is successful in inferring uncertainty, in terms of value/reward-over-belief calculation in time horizons, T=1 and T=2, by considering wireless agents' states (monitoring, emergency) and observations (EMI-effect and collision-avoidance). Proof of convergence and complexity analysis of RMSRA-QL-POSG are also presented.
展开▼