The system, called MLConvTrans — short for Multi-Level Convolutional Transformer — was developed by Thi Hong Vuong of VNU University of Engineering and Technology in Hanoi and the National Institute of Informatics in Tokyo, together with Tung Doan of Hanoi University of Science and Technology and Atsuhiro Takasu of the National Institute of Informatics and SOKENDAI. In benchmark tests spanning six widely used public datasets, the model delivered higher accuracy and F1-scores than a roster of state-of-the-art competitors, all while using fewer trainable parameters, fewer floating-point operations, and shorter inference times than other Transformer-based approaches.

Human activity recognition, or HAR, has expanded rapidly in recent years, finding roles in sports analytics, smart homes, and health monitoring, including fall detection, elderly care, and the tracking of abnormal movements in patients. The basic recipe is straightforward: accelerometers, gyroscopes, and sometimes magnetometers embedded in a wearable or handheld device capture three-axis motion signals, which software then segments into windows and classifies into activity labels such as walking, running, sitting, or jumping. What happens inside that classification stage, however, has been the subject of fierce competition among machine learning approaches.

Early HAR systems relied on classical algorithms such as Support Vector Machines, Decision Trees, and Random Forests. These methods demanded extensive hand-crafted feature engineering — human experts had to decide which statistical properties of the sensor signals mattered — and their accuracy and generalizability suffered as a result. Deep neural networks changed the game by learning features automatically from raw data, stacking layers that capture increasingly abstract patterns. Convolutional neural networks excelled at pulling local features out of multichannel sensor streams, but they struggled to model long-term temporal dependencies across extended sequences. Recurrent architectures, including Long Short-Term Memory networks and bidirectional variants, addressed temporal dynamics but often failed to capture correlations among heterogeneous sensor modalities. Hybrid CNN-LSTM models, which pair spatial feature extraction with temporal modeling, improved matters further yet still left long-range relationships and inter-modal physical correlations underexploited.

Transformers, the architecture family that underpins modern large language models, entered the HAR arena promisingly because their self-attention mechanism naturally captures long-range dependencies across a sequence. But as the authors note, existing Transformer-based HAR systems came with their own baggage. Original Transformer designs struggled to correlate isolated timestamps into cohesive activity representations, often required elaborate pre-processing pipelines to merge signals from multiple sensors — one method needed 273 engineered features extracted before the model could even run — and offered little in the way of efficiency analysis. Variants that combined CNNs with Transformers, such as IF-ConvTransformer and HART, partially alleviated these issues but introduced complexity of their own, whether through physically informed fusion blocks that limit sensor flexibility or through stacks of encoder layers that inflate inference time.

MLConvTrans attacks the problem with a two-stage design. The first component, a Multi-Level Convolutional Network dubbed MLConvNet, handles local feature extraction and multimodal fusion. It operates at two levels. At level one, a Conv Embedding block built from 1D convolutional layers with kernel size three, each followed by max pooling, processes the accelerometer and gyroscope streams separately, extracting sensor-specific local features before concatenating them into a unified embedding. At level two, these projected embeddings flow through three Residual Attention Blocks — convolutional layers with kernel size five, batch normalization, ReLU activation, and skip connections — which aggregate global, contextual information across the fused sensor data. The residual design keeps the network compact while enabling the extraction of complex spatial patterns, and the multilevel structure deliberately balances small kernels for fine-grained local detail with larger kernels for broader receptive fields.

The second component, the TransEncoder, is a stack of two Transformer encoder layers that receives MLConvNet’s output combined with sinusoidal positional encodings. Each encoder layer contains a multi-head self-attention block and a feed-forward multilayer perceptron with Gaussian Error Linear Unit activation. The self-attention mechanism computes queries, keys, and values from learned linear projections of the input, weighting each time step’s contribution to every other time step and thereby modeling long-term dependencies that convolutions alone cannot reach. A final classification block applies average pooling, batch normalization, a multilayer perceptron, and dropout, followed by a softmax that distributes probability across activity classes, trained with a cross-entropy loss.

The experimental campaign was unusually broad. The team evaluated MLConvTrans on UCI-HAR, MotionSense, HAPT, KU-HAR, SHL2018, and PAMAP2 — datasets collected from smartphone- and IMU-based setups covering anywhere from six to eighteen activities, sample rates from 50 to 100 Hz, and participants ranging from 9 to 90 per study. Raw accelerometer and gyroscope signals were channel-wise z-normalized, segmented into 128-timestamp windows, and split 70/10/20 into training, validation, and test sets. Training ran for 200 epochs with a batch size of 64 and the Adam optimizer at a learning rate of 0.0001 on Nvidia Tesla V100 GPUs, with class weighting to counter label imbalance, and every configuration was repeated ten times to produce averaged metrics.

The results were decisive. MLConvTrans achieved average F1-scores of 97.31 percent on UCI-HAR, 98.74 percent on MotionSense, 92.43 percent on HAPT, 95.93 percent on KU-HAR, 80.59 percent on SHL2018, and 96.76 percent on PAMAP2, with corresponding accuracies of 98.26, 98.87, 93.15, 96.82, 81.34, and 97.21 percent. Against CNN-based baselines such as DeepCNN, TCN, and AReNet, the F1-score gains ranged from roughly 4 to 19 percentage points depending on the dataset. Hybrid CNN-LSTM methods trailed by between 1.7 and 12.5 points, and even fellow Transformer approaches — the vanilla Transformer, IF-ConvTransformer, and HART — fell short by 3.1 to 7.8 points. Stratified five-fold cross-validation confirmed the model’s stability, holding above 95 percent accuracy on four of the six datasets.

Confusion-matrix analysis offered a nuanced picture of where errors persist. On UCI-HAR and MotionSense, most misclassifications occurred between standing and sitting — statically similar postures that even humans can confuse from motion data alone. HAPT proved harder because it includes postural transitions such as sit-to-lie and stand-to-lie, brief and unstable movements with few training samples. On SHL2018, a real-world transportation dataset, the model excelled at dynamic activities like walking, running, and cycling, but distinguishing among car, bus, train, and subway rides — where passengers mostly sit still — pushed individual class accuracies down to the 68 to 82 percent range, reflecting the intrinsic ambiguity of the signals rather than any architectural weakness.

Ablation studies confirmed that both components earn their place. Removing the TransEncoder and keeping only MLConvNet cost the model between 2.9 and 12.1 points of accuracy depending on the dataset, while removing MLConvNet and relying on the Transformer alone cost between 2.0 and 5.1 points. The team also tuned the architecture empirically: six convolutional layers paired with two Transformer layers proved the sweet spot, a kernel size of three in level one and five in level two outperformed uniform choices, and four attention heads hit the best accuracy-to-cost balance, with eight heads adding only marginal gains while inflating the parameter count by more than a quarter. Testing additional sensors showed that adding a magnetometer yields only modest improvements, suggesting the accelerometer-gyroscope pair remains the pragmatic choice for most deployments.

Perhaps most consequential for real-world use is the efficiency profile. MLConvTrans contains 1,576,924 trainable parameters and requires about 16.3 million FLOPs per inference — less than every Transformer-based competitor tested, including the vanilla Transformer at 2.01 million parameters and HART at 2.54 million. Its average inference time of 3.52 milliseconds was the lowest in the comparison, beating ConvTransformer, the vanilla Transformer, HART, and IF-ConvTransformer, the latter of which required nearly five milliseconds on the same hardware. That combination of accuracy and speed, the authors argue, makes the architecture genuinely suitable for real-time applications such as fall detection on battery-powered wearables.

The team’s future work aims to shrink the model further, building a lightweight HAR engine that can live comfortably inside smart devices with limited memory, storage, and energy budgets. If MLConvTrans’s benchmark performance holds outside the laboratory, the quiet motion sensors already strapped to millions of wrists and pockets may soon understand human movement considerably better — and understand it quickly enough to matter in the moment.

The Northern Goshawk Optimization algorithm takes its inspiration from the hunting strategies of the northern goshawk, a bird known for its precision and efficiency in capturing prey. This natural phenomenon is mirrored in the optimization technique, where the algorithm seeks to emulate the bird’s ability to make quick, efficient decisions based on environmental factors. By simulating these behaviors, the researchers have designed a model that can dynamically adjust its parameters to enhance battery health predictions, making it an exciting advancement in the realm of artificial intelligence applied to energy storage systems.

Lithium-ion batteries, while ubiquitous in today’s technology, present significant challenges in predictive maintenance and health monitoring. Traditional methods rely heavily on static models and historical data, often leading to inaccurate estimations of battery life and performance. The introduction of the hybrid neural network aims to address these shortcomings by leveraging deep learning capabilities to learn from a continuous influx of real-time data. This means that as the battery operates, the neural network adapts and learns, providing a highly responsive and accurate assessment of the battery’s state of health.

One of the most remarkable aspects of this new approach is its ability to handle complex datasets which include variables such as temperature, charge cycles, and voltage fluctuations. The integration of the Northern Goshawk Optimization algorithm with the neural network allows the system to prioritize and weigh these various data points effectively. As a result, the algorithm can quickly identify patterns and anomalies that could indicate potential issues, leading to timely interventions that can prolong battery life and enhance overall performance.

In addition to improving battery health estimations, the implications of this research stretch far beyond just battery management. The methods developed in this study could be applied to a variety of other fields that rely on predictive modeling and optimization. Industries such as renewable energy, electric vehicles, and grid management could greatly benefit from the enhanced accuracy of health monitoring systems, driving efficiency and reliability in these crucial sectors.

The researchers conducted extensive experiments comparing the performance of their hybrid model against existing traditional methods. The results were astonishing, showcasing a marked improvement in estimation accuracy. This was achieved not only through the synergistic blending of optimization techniques and machine learning but also through meticulous validation of the model against real-world data acquired from operational batteries.

Furthermore, the energy sector is at a pivotal juncture, with a growing emphasis on sustainability and reducing carbon footprints. Enhancements in battery technology are essential for the wider adoption of electric vehicles and renewable energy sources. The model presented by Zhang et al. directly addresses these challenges, providing stakeholders with the tools necessary to ensure the longevity and reliability of lithium-ion batteries, thereby facilitating a smoother transition to a more sustainable energy future.

While the technical intricacies of the model can be challenging to grasp, the essence lies in its capacity for ongoing learning and adaptation. This characteristic is crucial as the landscape of battery technology continues to evolve rapidly. As electric vehicles become more commonplace and renewable energy sources increase their share in global energy production, the need for reliable battery health monitoring systems will be greater than ever.

The trajectory of this research is promising, and the scientific community is eagerly awaiting further developments and practical implementations of this innovative model. As researchers continue to refine the algorithms and expand their applications, the collaboration between natural phenomena, artificial intelligence, and energy technology stands as a beacon of potential advancements.

The integration of artificial intelligence into battery management systems raises important discussions surrounding data privacy and cybersecurity. As these systems become more interconnected, the data they process becomes invaluable. Ensuring that the information is secure and protected against potential threats becomes paramount. The researchers recognize that while the technical advancements in battery health estimation are groundbreaking, addressing ethical considerations around data usage is equally important for the trust and safety of users.

In conclusion, the innovative northern goshawk optimization – hybrid neural network algorithm heralds a new chapter in the field of lithium-ion battery technology. Its promise for highly accurate health estimation has critical implications for various high-stakes industries. As the world continues its shift towards new energy solutions, this research not only underscores a significant leap in technological capability but also highlights the ongoing partnership between nature and science in solving contemporary challenges.

As the release date of this research approaches, anticipation builds among researchers and industry leaders alike. The potential applications and benefits of this technology could reshape how we understand and manage one of the most pivotal components of modern electronics and eco-friendly solutions. The discourse surrounding the advancements will undoubtedly contribute to a broader understanding of the role that advanced optimization algorithms and neural networks play in the evolution of energy systems.

Subject of Research: State of health estimation of lithium-ion batteries using hybrid neural network and optimization algorithms.

Article Title: An innovative northern goshawk optimization – hybrid neural network algorithm for highly accurate state of health estimation of lithium-ion batteries.

Article References:

Zhang, L., Liu, D., Wang, S. et al. An innovative northern goshawk optimization – hybrid neural network algorithm for highly accurate state of health estimation of lithium-ion batteries.
Ionics (2025). https://doi.org/10.1007/s11581-025-06836-7

Image Credits: AI Generated

DOI: https://doi.org/10.1007/s11581-025-06836-7

Keywords: lithium-ion batteries, state of health estimation, Northern Goshawk Optimization, hybrid neural network, predictive maintenance, artificial intelligence, energy technology.