Current trends in sensing call for continuous, pervasive and unobtrusive operation under constrained and highly complex environments. Autonomic sensing addresses many of the related practical requirements, empowering both the ability to minimize human intervention as well as to achieve resilient data delivery. Driven by the necessity for distributed operation, in this talk we will explore network techniques that benefit the practical realization of autonomic sensing. Serving as the cornerstone of the network backbone, our emphasis will be given on connectivity issues, examined by two different, yet complementary, aspects: (a) the topology control that enables guarantees for end-to-end connectivity, (b) the topology reconfiguration against varying operational spaces. The common characteristic of the presented approaches is the reliance on the information that each node can capture, process and interpret, without the need to generate overhead or obtain an overall perspective of the global network conditions. Rigorous simulation is coupled to experimental results for examining the efficacy of the proposed approaches, whilst future directions that enable a tighter alignment to on-node processing are open for discussion.