With the rapid growth of the Internet of Things (IoT), integrating artificial intelligence (AI) on extremely weak embedded devices has garnered significant attention, enabling improved real-time performance and enhanced data privacy. However, the resource limitations of such devices and unreliable network conditions necessitate error-resilient device-edge collaboration systems. Traditional approaches focus on bit-level transmission correctness, which can be inefficient under dynamic channel conditions. In contrast, we propose SemanticNN, a semantic codec that tolerates bit-level errors in pursuit of semantic-level correctness, enabling compressive and resilient collaborative inference offloading under strict computational and communication constraints. It incorporates a Bit Error Rate (BER)-aware decoder that adapts to dynamic channel conditions and a Soft Quantization (SQ)-based encoder to learn compact representations. Building on this architecture, we introduce Feature-augmentation Learning, a novel training strategy that enhances offloading efficiency. To address encoder-decoder capability mismatches from asymmetric resources, we propose XAI-based Asymmetry Compensation to enhance decoding semantic fidelity. We conduct extensive experiments on STM32 using three models and six datasets across image classification and object detection tasks. Experimental results demonstrate that, under varying transmission error rates, SemanticNN significantly reduces feature transmission volume by 56.82-344.83x while maintaining superior inference accuracy.

翻译：随着物联网（IoT）的快速发展，在极端弱计算能力的嵌入式设备上集成人工智能（AI）技术受到广泛关注，这有助于提升实时性能并增强数据隐私保护。然而，此类设备的资源限制及不可靠的网络条件要求系统具备容错能力的设备-边缘协同机制。传统方法侧重于比特级传输的正确性，在动态信道条件下可能效率低下。相比之下，我们提出了SemanticNN——一种语义编解码器，它容忍比特级错误以追求语义级正确性，从而在严格的计算与通信约束下实现压缩且鲁棒的协同推理卸载。该框架包含一个感知比特错误率（BER）的自适应动态信道解码器，以及基于软量化（SQ）的编码器以学习紧凑表示。基于此架构，我们引入了特征增强学习这一新颖训练策略，以提升卸载效率。为应对因资源不对称导致的编码器-解码器能力失配，我们提出基于可解释人工智能（XAI）的非对称补偿机制，以增强解码语义保真度。我们在STM32平台上使用三种模型和六个数据集，针对图像分类与目标检测任务进行了大量实验。实验结果表明，在不同传输错误率下，SemanticNN能将特征传输量显著降低56.82-344.83倍，同时保持优异的推理精度。