Weather-Driven Perception for Autonomous Driving and Remote Sensing with Spiking Neural Networks: Concepts, Methods, and Perspectives
Adverse weather is a major challenge for reliable and energy-efficient autonomous driving because it degrades sensing and propagates uncertainty through perception, planning, and control. Spiking neural networks (SNNs) are promising for vehicle edge systems because their event-driven computation can reduce latency, memory use, and energy consumption. Their effectiveness, however, depends on how weather changes sensor inputs and encoded spike streams. Existing SNN surveys rarely address adverse weather, while weather-focused perception surveys seldom examine spike-based processing. This survey bridges these gaps through a weather-first framework covering event cameras, frame cameras, LiDAR, radar, and synthetic aperture radar, from sensing and spike encoding to neuron dynamics, learning, multimodal fusion, and neuromorphic hardware. We formalise weather degradation as evidence attenuation and non-semantic activity, and synthesise findings from public datasets with real, emulated, simulated, and synthetic results explicitly distinguished. We review mitigation strategies including adaptive encoding, denoising, weather-diverse training, reliability-aware fusion, and hardware-aware spike control. We formulate five testable hypotheses; identify open problems in real-weather benchmarks, cross-modal evaluation, measured energy, and neuromorphic deployment; and outline future research directions for robust, energy-efficient SNN-based autonomous driving.
Authors
- James Ndodana Njaji (ORCID: https://orcid.org/0000-0002-0693-2015)
- Alexander Carballo (ORCID: https://orcid.org/0000-0002-5941-2195)
Institutions
- Gifu University (JP)
Publication Details
- Journal
- Sensors
- Published
- 2026-11-20
- DOI
- https://doi.org/10.3390/s26227348
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Japan Science and Technology Agency