Knowledge-based design space exploration of wireless sensor networks

The complexity of Wireless Sensor Networks (WSNs) has been constantly increasing over the last decade, and the necessity of efficient CAD tools has been growing accordingly. In fact, the size of the design space of a WSN has become large, and an exploration conducted by using semi-random algorithms (such as the popular genetic or simulated annealing algorithms) requires an unacceptable amount of time to converge due to the high number of parameters involved. To address this issue, in this paper we introduce a knowledge-based design space exploration algorithm for the WSN domain, which is based on a discrete-space Markov decision process (MDP). In order to enhance the performance of the proposed algorithm and to increase its scalability, we tailor the classical MDP approach to the specific aspects that characterize the WSN domain. We exploit domain-specific knowledge to choose the best node-level configuration in WSNs using slotted star topology in order to reduce the exploration time. The proposed approach has been tested on IEEE 802.15.4 star networks with various configurations of the number of nodes and their packet rates. Experimental results show that the proposed algorithm reduces the number of simulations required to converge, with respect to state-of-the-art algorithms (e.g., NSGA-II, PMA and MOSA), from 60 to 87%.