EdgeSynapse is a mechanism for energy-efficient sensing in remote, energy-constrained settings. It leverages neuromorphic principles by having sensor nodes and cluster heads generate discrete spikes when local integrator states cross thresholds. Specifically, each sensor uses a leaky integrate-and-fire (LIF) model to accumulate observations. When the membrane potential exceeds a threshold, the node emits an excitatory spike via LoRaWAN uplink and resets. Cluster heads similarly integrate incoming spikes and forward a higher-level spike when their threshold is reached, optionally using LoRaWAN downlink slots for inhibitory spikes to moderate traffic. This hierarchical spiking approach mimics biological signaling and reduces unnecessary transmissions. We provide a mathematical model of the LIF integration at sensors and cluster heads, along with the transmission algorithm tailored to LoRaWAN Class A communication. A preliminary simulation using real sensor data reveals favorable energy-accuracy trade-offs and a significant reduction in transmissions relative to send-on-Delta, alongside improved stability under synchronized bursts.

Top