We propose EdgeSpike, a co-designed spiking neural network (SNN) framework for autonomous low-power sensing in edge Internet of Things (IoT) architectures. EdgeSpike unifies (i) a hybrid surrogate-gradient and direct-encoding training pipeline, (ii) a hardware-aware neural architecture search (NAS) bounded by per-inference energy and memory budgets, (iii) an event-driven runtime targeting Intel Loihi 2, SpiNNaker 2, and commodity ARM Cortex-M microcontrollers with custom spike-sparse SIMD kernels, and (iv) a lightweight local plasticity rule enabling continual on-device adaptation without backpropagation. The framework is evaluated across five sensing tasks (keyword spotting, vibration-based machine fault detection, surface electromyography gesture recognition, 77 GHz radar human-activity classification, and structural-health acoustic-emission monitoring) on three hardware targets. EdgeSpike achieves a mean classification accuracy of 91.4%, within 1.2 percentage points (pp) of strong INT8 convolutional neural network (CNN) baselines (mean 92.6%), while reducing energy per inference by 18x to 47x on neuromorphic hardware (mean 31x) and by 4.6x to 7.9x on Cortex-M (mean 6.1x). End-to-end latency remains at or below 9.4 ms across all 15 task-hardware configurations. A seven-month, 64-node wireless field deployment confirms a 6.3x extension in projected battery lifetime (from 312 to 1978 days at 2 Wh per node) and bounded accuracy degradation under seasonal drift (0.7 pp with on-device adaptation versus 2.1 pp without). Hardware-aware NAS evaluates 8400 candidates and yields a 12-point Pareto front. EdgeSpike will be released as open source with reproducible training pipelines, hardware-portable runtimes, and benchmark suites.

翻译：我们提出 EdgeSpike，一种面向边缘物联网（IoT）架构中自主低功耗感知的协同设计脉冲神经网络（SNN）框架。EdgeSpike 统一了以下四个方面：（i）混合代理梯度与直接编码训练流程；（ii）受每次推理能量与内存预算约束的硬件感知神经架构搜索（NAS）；（iii）面向 Intel Loihi 2、SpiNNaker 2 及商用 ARM Cortex-M 微控制器、配备自定义脉冲稀疏单指令多数据（SIMD）内核的事件驱动运行时；（iv）一种轻量级局部可塑性规则，可实现无反向传播的持续设备端自适应。该框架在三个硬件目标上，针对五项感知任务（关键词识别、基于振动的机器故障检测、表面肌电图手势识别、77 GHz 雷达人体活动分类及结构健康声发射监测）进行了评估。EdgeSpike 取得了 91.4% 的平均分类准确率，与强 INT8 卷积神经网络（CNN）基线（平均 92.6%）相比仅差 1.2 个百分点（pp），同时在神经形态硬件上将每次推理能耗降低了 18 倍至 47 倍（平均 31 倍），在 Cortex-M 上降低了 4.6 倍至 7.9 倍（平均 6.1 倍）。在所有 15 种任务-硬件配置下，端到端延迟保持在 9.4 毫秒或以下。为期七个月、包含 64 个节点的无线现场部署证实，预计电池寿命延长了 6.3 倍（每个节点 2 瓦时下，从 312 天延长至 1978 天），且在季节性漂移下准确率退化有限（使用设备端自适应时为 0.7 pp，而未使用时为 2.1 pp）。硬件感知 NAS 评估了 8400 个候选架构，生成了 12 点的帕累托前沿。EdgeSpike 将以开源形式发布，包含可复现的训练流程、硬件可移植运行时及基准测试套件。