"UAV-SNN Review Dataset"

"The third-generation, neurologically-inspired Spiking Neural Networks (SNNs) possess many attributeswhich make them attractive to the often restrictive operating environment found within Unmanned AerialVehicles (UAVs). These benefits include, but are not limited to, high power efficiency, low latency, datasparsity, event-driven processing, synaptic plasticity and neuromorphic hardware compatibility. In thissurvey of 34 studies spanning the period 2019-2025 which implemented spiking networks onboard a real orsimulated UAV, we identified which models, methods, and implementations were frequently utilised, andwhich were under-explored. Specifically, for each study, we investigated the questions of which UAV-relatedtasks were solved by the system, which role the SNNs played in the pipeline, what neuron models, trainingstrategies and spike coding methods were employed by SNNs, and which sensory inputs, model sizes andneuromorphic processing chips were utilised by UAVs. We found that, in terms of task performance alone,some studies reported SNNs to perform better than traditional counterparts in UAV-related tasks, whilemost found no such difference. Yet SNNs bring more to the table than performance alone, including theaforementioned advantages. Nevertheless, more work needs to be done to fully utilise these potentials withinUAVs, including developing more scalable SNNs, integrating into neuromorphic hardware more efficiently,utilising the temporal information within spikes more effectively, standardisation of SNN encoding methodsand benchmarks, online learning plasticity and applications on more diverse types of UAV."