A heterogeneous system architecture for low-power wireless sensor nodes in compute-intensive distributed applications

Wireless Sensor Networks (WSNs) are a key technology for future social and industrial developments. With increasingly complex applications, the compute demands for in-node / in-network processing have been growing steadily, while the capabilities of energy-harvesting or -storage systems have advanced only slowly. We present the Hardware-Accelerated Low Power Mote (HaLoMote), a heterogeneous system architecture for a wireless sensor node that achieves significantly better energy efficiency than traditional approaches, even for demanding applications requiring sensor sample rates of hundreds of Hertz. The paper discusses an evolution of the node hardware architecture and details the implementation of a signal processing chain required for Structural Health Monitoring (SHM) applications. The measurement accuracy of the WSN-based data acquisition system is compared against a wire-bound laboratory system showing that the dominant eigenfrequencies of a monitored structure can be detected with less than 1% error. Furthermore, the runtime and power requirements of the HaLoMote are compared against various software processors typically used for conventional WSN architectures. It can be shown that the HaLoMote is 2.3 times more energy-efficient than a state-of-the-art ARM Cortex-M3 based device.