As autonomous driving (AD) systems grow more complex, their rising computational demands pose significant energy and sustainability challenges. This paper investigates spiking neural networks (SNNs) as low‐power alternatives to convolutional neural networks (CNNs) for regression tasks in AD. We introduce a membrane‐potential ( V _mem ) decoding framework that converts binary spike trains into continuous outputs and propose the energy‐to‐error ratio (EER), a unified metric combining prediction error with energy consumption. Three CNN architectures (PilotNet, LaksNet, and MiniNet) and their corresponding SNN variants are trained and evaluated using delta, latency, and rate encoding across varied parameter settings, with energy use and emissions logged. Delta‐encoded SNNs achieve the highest EER, substantial energy savings with minimal performance loss, whereas CNNs, despite slightly better MSE, incur 10–20 × higher energy costs. Rate encoding underperforms, and latency encoding, though improving relative error, demands excessive energy. Parameter tuning (threshold θ , temporal dynamics ( S ), membrane time constant ( τ ), and gain G ) directly influences eco‐efficiency. All experiments run on standard GPUs, showing SNNs can surpass CNNs in eco‐efficiency without specialized hardware. Paired statistical tests confirm that only delta‐encoded SNNs achieve significant EER improvements. This work presents a practical, energy‐aware evaluation framework for neural architectures, establishing EER as a critical metric for sustainable machine learning in intelligent transport and beyond.

This letter presents a practical and energy-aware framework for deploying Spiking Neural Networks on low-cost hardware for edge computing on existing software and hardware components. We detail a reproducible pipeline that integrates neuromorphic processing with secure remote access and distributed intelligence. Using Raspberry Pi and the BrainChip Akida PCIe accelerator, we demonstrate a lightweight deployment process including model training, quantization, and conversion. Our experiments validate the eco-efficiency and networking potential of neuromorphic AI systems, providing key insights for sustainable distributed intelligence. This letter offers a blueprint for scalable and secure neuromorphic deployments across edge networks, highlighting the novelty of providing a reproducible integration pipeline that brings together existing components into a practical, energy-efficient framework for real-world use.

This work explores the application of Spiking Neural Networks (SNNs) and Continual Learning (CL) methodologies to the problem of steering angle regression, using autonomous driving simulation as the experimental context, with a focus on energy efficiency and alignment with sustainable computing objectives. The primary goal was to design and implement CL techniques in SNNs to assess the model’s ability to maintain accuracy in explored environments while reducing CO₂ emissions through the optimized use of a subset of the data. This study emerges in response to the increasing energy demand of deep learning models, which poses a challenge to sustainability. SNNs, inspired by the efficiency of biological neural systems, offer significant advantages in terms of computational and energy consumption, making them a promising alternative. CL techniques, such as Elastic Weight Consolidation and replay memory, are integrated to mitigate catastrophic forgetting in sequential learning tasks. The methodology includes adapting the PilotNet architecture for SNNs, preprocessing datasets generated in the Udacity driving simulator, and evaluating models in incremental learning scenarios. The experiments compare the performance of SNNs with CL against baseline models without CL, using mean squared error (MSE), computational efficiency, and equivalent CO₂ emissions as evaluation metrics. The results demonstrate that replay memory enables the retention of prior knowledge with a limited increase in energy consumption. This work concludes that SNNs with CL are a viable alternative for sustainable AI applications. Future research directions include a focus primarily on hardware-specific implementations and real-world testing.

Although the spotlight on ecological efficiency in artificial intelligence (AI) research is intensifying, its integration remains underdeveloped. Addressing this gap is essential to align algorithmic performance with ecological responsibility and support the carbon neutrality targets of the Organization for Economic Cooperation and Development (OECD). This study proposes a methodology to measure CO ₂ emissions and energy consumption during dataset encoding and training of Spiking Neural Networks (SNNs) in regression tasks. Using rate encoding with varied hyperparameters, we compare the original PilotNet convolutional neural network (CNN) with its SNN adaptation across three autonomous driving datasets: Udacity, Sully Chen (Palo Alto), and AirSim. Results show that temporal depth ( S , number of time steps) is the primary driver of environmental footprint, with increases from S = 5 to S = 25 raising emissions without proportional accuracy gains, while the gain parameter ( G , scaling factor for spike generation probability) has negligible impact. On Udacity, CNN outperforms SNNs in both accuracy and footprint. On Palo Alto, an SNN slightly improves accuracy but at significantly higher emissions. On AirSim, SNNs outperform CNN in accuracy but incur substantially greater environmental cost. These findings indicate that, for rate‐encoded SNNs of this architecture executed on conventional CPU/GPU hardware, the accuracy‐energy trade‐off varies across driving environments and that ecoefficiency gains remain limited, requiring neuromorphic platforms and improved metrics capturing both performance and environmental impact.