Limited Computational Resources Research Articles

XimSwap: Many-to-Many Face Swapping for TinyML

The unprecedented development of deep learning approaches for video processing has caused growing privacy concerns. To ensure data analysis while maintaining privacy, it is essential to address how to protect individuals’ identities. One solution is to anonymize data at the source, avoiding the transmission or storage of information that could lead to identification. This study introduces XimSwap, a novel deep learning technique for real-time video anonymization, which can remove facial identification features directly on edge devices with minimal computational resources. Our approach offers a comprehensive solution that guarantees privacy by design. This novel method for implementing face-swapping ensures that the pose and expression of a target face remain unchanged and can be used on embedded devices with very limited computational resources. By incorporating style transfer layers into convolutional ones and optimizing the network’s operation, we achieved a reduction of over 98% in the required operations and parameters compared with state-of-the-art architectures. Our approach also significantly reduces RAM usage, making it possible to implement the anonymization process on tiny edge devices, including microcontrollers, such as the STM32H743.

Internet of Things intrusion detection: Research and practice of NSENet and LSTM fusion models

To address the problems of complex environment, limited device computational resources and limited memory resources in the existing IoT, SELSTM, an intrusion detection system composed of NSENet and LSTM fusion based on SENet, is investigated. The NSENet part of the SELSTM system is based on the squeeze-and-excitation network (SENet). The lightweight computational modules NonLocal, SKConv and inverted residuals are fused into SE blocks, and self-attention of Nonlocal is used to improve the local receptive field of feature extraction. The channel attention and spatial attention of each part of the data are strengthened by the use of SKConv To enhance the adaptive convolution ability of the model and ensure the completeness of the information, the properties of the inverted residual structure are used to ensure that the gradient of the model decreases steadily without gradient explosion or disappearance. For the problem of data imbalance, the dataset is randomly resampled using the weight resampling technique to improve the balance of the dataset to ensure that the final detection effect of the model is more effective and generalized, while the data flow is divided into two parts for processing, and the model parameters are optimized using the model gradient optimizer consisting of the optimizer Lion and the optimization function Lookahead. The model extracts the spatial and temporal features of the data through multidimensional extraction to ensure the completeness of the data feature information in multiple dimensions, thus obtaining better detection results. The results of the experiments comparing the SELSTM model with other models on the intrusion dataset show that the intrusion detection model has a higher detection precision and accuracy than the traditional deep learning intrusion detection model, which indicates that the SELSTM has better detection performance properties and better practicality and effectiveness on IoT devices.

Quantized non-volatile nanomagnetic domain wall synapse based autoencoder for efficient unsupervised network anomaly detection

Abstract Anomaly detection in real-time using autoencoders implemented on edge devices is exceedingly challenging due to limited hardware, energy, and computational resources. We show that these limitations can be addressed by designing an autoencoder with low-resolution non-volatile memory-based synapses and employing an effective quantized neural network learning algorithm. We further propose nanoscale ferromagnetic racetracks with engineered notches hosting magnetic domain walls (DW) as exemplary non-volatile memory based autoencoder synapses, where limited state (5-state) synaptic weights are manipulated by spin orbit torque (SOT) current pulses to write different magnetoresistance states. The performance of anomaly detection of the proposed autoencoder model is evaluated on the NSL-KDD dataset. Limited resolution and DW device stochasticity aware training of the autoencoder is performed, which yields comparable anomaly detection performance to the autoencoder having floating-point precision weights. While the limited number of quantized states and the inherent stochastic nature of DW synaptic weights in nanoscale devices are typically known to negatively impact the performance, our hardware-aware training algorithm is shown to leverage these imperfect device characteristics to generate an improvement in anomaly detection accuracy (90.98%) compared to accuracy obtained with floating-point synaptic weights that are extremely memory intensive. Furthermore, our DW-based approach demonstrates a remarkable reduction of at least three orders of magnitude in weight updates during training compared to the floating-point approach, implying significant reduction in operation energy for our method. This work could stimulate the development of extremely energy efficient non-volatile multi-state synapse-based processors that can perform real-time training and inference on the edge with unsupervised data.

Adaptive Graph Convolutional Recurrent Network with Transformer and Whale Optimization Algorithm for Traffic Flow Prediction

Accurate traffic flow prediction plays a crucial role in the development of intelligent traffic management. Despite numerous investigations into spatio-temporal methods, achieving high accuracy in traffic flow prediction remains challenging. This challenge arises from the complex dynamic spatio-temporal correlations within the traffic road network and the limitations imposed by the selection of hyperparameters based on experiments and manual experience, which can affect the performance of the network architecture. This paper introduces a novel transformer-based adaptive graph convolutional recurrent network. The proposed network automatically infers the interdependencies among different traffic sequences and incorporates the capability to capture global spatio-temporal correlations. This enables the dynamic capture of long-range temporal correlations. Furthermore, the whale optimization algorithm is employed to efficiently design an optimal network structure that aligns with the requirements of the traffic domain and maximizes the utilization of limited computational resources. This design approach significantly enhances the model’s performance and improves the accuracy of traffic flow prediction. The experimental results on four real datasets demonstrate the efficacy of our approach. In PEMS03, it improves MAE by 2.6% and RMSE by 1.4%. In PEMS04, improvements are 1.6% in MAE and 1.4% in RMSE, with a similar MAPE score to the best baseline. For PEMS07, our approach shows a 4.1% improvement in MAE and 2.2% in RMSE. On PEMS08, it surpasses the current best baseline with a 3.4% improvement in MAE and 1.6% in RMSE. These results confirm the good performance of our model in traffic flow prediction across multiple datasets.

Adaptive Multi-Criteria Selection for Efficient Resource Allocation in Frugal Heterogeneous Hadoop Clusters

Efficient resource allocation is crucial in clusters with frugal Single-Board Computers (SBCs) possessing limited computational resources. These clusters are increasingly being deployed in edge computing environments in resource-constrained settings where energy efficiency and cost-effectiveness are paramount. A major challenge in Hadoop scheduling is load balancing, as frugal nodes within the cluster can become overwhelmed, resulting in degraded performance and frequent occurrences of out-of-memory errors, ultimately leading to job failures. In this study, we introduce an Adaptive Multi-criteria Selection for Efficient Resource Allocation (AMS-ERA) in Frugal Heterogeneous Hadoop Clusters. Our criterion considers CPU, memory, and disk requirements for jobs and aligns the requirements with available resources in the cluster for optimal resource allocation. To validate our approach, we deploy a heterogeneous SBC-based cluster consisting of 11 SBC nodes and conduct several experiments to evaluate the performance using Hadoop wordcount and terasort benchmark for various workload settings. The results are compared to the Hadoop-Fair, FOG, and IDaPS scheduling strategies. Our results demonstrate a significant improvement in performance with the proposed AMS-ERA, reducing execution time by 27.2%, 17.4%, and 7.6%, respectively, using terasort and wordcount benchmarks.

Continuous lipreading based on acoustic temporal alignments

Visual speech recognition (VSR) is a challenging task that has received increasing interest during the last few decades. Current state of the art employs powerful end-to-end architectures based on deep learning which depend on large amounts of data and high computational resources for their estimation. We address the task of VSR for data scarcity scenarios with limited computational resources by using traditional approaches based on hidden Markov models. We present a novel learning strategy that employs information obtained from previous acoustic temporal alignments to improve the visual system performance. Furthermore, we studied multiple visual speech representations and how image resolution or frame rate affect its performance. All these experiments were conducted on the limited data VLRF corpus, a database which offers an audio-visual support to address continuous speech recognition in Spanish. The results show that our approach significantly outperforms the best results achieved on the task to date.

Ensemble learning for retinal disease recognition under limited resources.

Retinal optical coherence tomography (OCT) images provide crucial insights into the health of the posterior ocular segment. Therefore, the advancement of automated image analysis methods is imperative to equip clinicians and researchers with quantitative data, thereby facilitating informed decision-making. The application of deep learning (DL)-based approaches has gained extensive traction for executing these analysis tasks, demonstrating remarkable performance compared to labor-intensive manual analyses. However, the acquisition of retinal OCT images often presents challenges stemming from privacy concerns and the resource-intensive labeling procedures, which contradicts the prevailing notion that DL models necessitate substantial data volumes for achieving superior performance. Moreover, limitations in available computational resources constrain the progress of high-performance medical artificial intelligence, particularly in less developed regions and countries. This paper introduces a novel ensemble learning mechanism designed for recognizing retinal diseases under limited resources (e.g., data, computation). The mechanism leverages insights from multiple pre-trained models, facilitating the transfer and adaptation of their knowledge to retinal OCT images. This approach establishes a robust model even when confronted with limited labeled data, eliminating the need for an extensive array of parameters, as required in learning from scratch. Comprehensive experimentation on real-world datasets demonstrates that the ensemble models constructed by the proposed ensemble method show superior performance over the baseline models under sparse labeled data, especially the triple ensemble model, which achieves the accuracy of 92.06%, which is 8.27%, 7.99%, and 11.14% better than the three baseline models, respectively. In addition, compared with the three baseline models learned from scratch, the triple ensemble model has fewer trainable parameters, only 3.677M, which is lower than the three baseline models of 8.013M, 4.302M, and 20.158M, respectively.

A relationship-aware mutual learning method for lightweight skin lesion classification

In recent years, deep learning has made significant advancements in skin cancer diagnosis. However, most methods prioritize high prediction accuracy without considering the limitations of computational resources, making them impractical for wearable devices. In this case, knowledge distillation has emerged as an effective method, capable of significantly reducing a model's reliance on computational and storage resources. Nonetheless, previous research suffers from two limitations: 1) the student model can only passively receive knowledge from the teacher model, and 2) the teacher model does not effectively model sample relationships during training, potentially hindering the effective transfer of sample relationship-related knowledge during knowledge distillation. To address these issues, we employ two identical student models, each equipped with a sample relationship module. This design ensures that the student models can mutually learn while modeling sample relationships. We conducted extensive experiments on the ISIC 2019 dataset to validate the effectiveness of our method. The results demonstrate that our approach significantly improves the recognition of various types of skin diseases. Compared to state-of-the-art methods, our approach exhibits higher accuracy and better generalization capabilities.

A two-stage direction-guided evolutionary algorithm for large-scale multiobjective optimization

Large-scale multiobjective optimization problems (LSMOPs) have exponential growth in the search space as the decision variables increase, and the vast search space poses a challenge to the performance of multiobjective evolutionary algorithms (MOEAs). Many current large-scale MOEAs need to consume a large amount of computational resources to get good performance. This paper proposes a two-stage direction-guided evolutionary algorithm for large-scale multiobjective optimization (LMOEA-S2D) to balance the performance and computational resource overhead. The algorithm exploits the Pareto-optimality property of domination and the diversity-preserving property of decomposition to optimize the performance in the two stages, respectively, and designs a corresponding direction-guided mechanism to improve search efficiency. LMOEA-S2D designs global direction search and local direction search in the domination-based stage for efficient exploitation to accelerate population convergence. To promote greater population diversity, a hybrid direction search was devised to aid diversity exploration in the decomposition-based stage, and this facilitates even distribution of candidate solutions across the Pareto optimal frontier. LMOEA-S2D is compared with five state-of-the-art large-scale MOEAs on some large-scale multiobjective test suites with 100 to 5,000 decision variables. The experimental results show that LMOEA-S2D significantly outperformed all compared algorithms under limited computational resources.

End-to-end automated speech recognition using a character based small scale transformer architecture

This study explores the feasibility of constructing a small-scale speech recognition system capable of competing with larger, modern automated speech recognition (ASR) systems in both performance and word error rate (WER). Our central hypothesis posits that a compact transformer-based ASR model can yield comparable results, specifically in terms of WER, to traditional ASR models while challenging contemporary ASR systems that boast significantly larger computational sizes. The aim is to extend ASR capabilities to under-resourced languages with limited corpora, catering to scenarios where practitioners face constraints in both data availability and computational resources. The model, comprising a compact convolutional neural network (CNN) and transformer architecture with 2.214 million parameters, challenges the conventional wisdom that large-scale transformer-based ASR systems are essential for achieving high accuracy. In comparison, contemporary ASR systems often deploy over 300 million parameters. Trained on a modest dataset of approximately 3000 h—significantly less than the 50,000 h used in larger systems—the proposed model leverages the Common Voice and LibriSpeech datasets. Evaluation on the LibriSpeech test-clean and test-other datasets produced character error rates (CERs) of 6.40% and 16.73% and WERs of 16.03% and 35.51% respectively. Comparisons with existing architectures showcase the efficiency of our model. A gated recurrent unit (GRU) architecture, albeit achieving lower error rates, incurred a computational cost 24 times larger than our proposed model. Large-scale transformer architectures, while achieving marginally lower WERs (2%–4% on LibriSpeech test-clean), require 200 times more parameters and 53,000 additional hours of training data. Modern large language models are used to improve the WERs, but require large computational resources. To further enhance performance, a small 4-g language model was integrated into our end-to-end ASR model, resulting in improved WERs. The overarching goal of this work is to provide a practical solution for practitioners dealing with limited datasets and computational resources, particularly in the context of under-resourced languages.

A risk-aware and recommender distributed intrusion detection system for home robots

The increasing adoption of home robots, driven by technological advancements and realistic goals, underscores the critical importance of ensuring their security, safety, and privacy. An Intrusion Detection System (IDS) is vital in safeguarding home robots by identifying and responding to potential security threats. However, home robot controllers’ limited computational and energy resources pose challenges. To enhance a home robot’s performance, especially its IDS, one potential solution is to connect it to a home network with a more robust computing infrastructure. Our proposed IDS comprises two components: one resides on a remote PC equipped with substantial computing power, and the other operates directly within the robot controller. By leveraging the remote PC, we achieve enhanced intrusion detection performance and gain valuable recommendations for fine-tuning the intrusion detection component situated in the controller. Simultaneously, it becomes imperative to swiftly identify security threats and prioritize high-risk attacks within the robot controller In this study, we employed machine learning techniques within the IDS. The dataset was generated by simulating robot actions within a home network. Based on our evaluation results, adopting a deep learning model for detecting intrusions in a remote PC and with its recommendations for fine-tuning the intrusion detector in the controller would significantly enhance intrusion detection performance. Additionally, leveraging ensemble learning to identify intrusions in the robot controller component while carefully considering potential risks promises satisfactory outcomes in terms of intrusion detection.

An Improved Body Force Model to Simulate Blade Loading Variation Under Various Operating Conditions in Transonic Axial Flow Compressor

Abstract In order to analyze the unsteady aerodynamic response of compressors with limited computational resources, the development of the body force (BF) model has been pursued for decades. This model can simulate the effects of the compressor blade on the flow by applying a three-dimensional (3D) force field. Indeed, the accuracy of the BF model simulation highly depends on the construction strategy of BF field. However, for a transonic compressor, the influence of shock waves on the spatial distribution of blade loading is significant, which makes the construction of an appropriate BF field challenging. To solve this issue, an accurate BF model is proposed to describe the effects of the blade on the flow with high spatial fidelity in transonic compressors. For verification, a typical transonic fan rotor, NASA-Rotor 67, is selected. The numerical results calculated by the new BF model are compared with those obtained by the Reynolds-averaged Navier–Stokes (RANS) solver. According to the results, the new model can accurately capture the distribution of blade loading in the transonic blade channel at various operating conditions, eventually leading to a good prediction of compressor performance and the radial distribution of flow parameters downstream of the rotor. Moreover, this study discusses the causes of modeling errors and presents the application of the model in inlet distortion simulation at the end.

A Lightweight Model of Underwater Object Detection Based on YOLOv8n for an Edge Computing Platform

The visual signal object detection technology of deep learning, as a high-precision perception technology, can be adopted in various image analysis applications, and it has important application prospects in the utilization and protection of marine biological resources. While the marine environment is generally far from cities where the rich computing power in cities cannot be utilized, deploying models on mobile edge devices is an efficient solution. However, because of computing resource limitations on edge devices, the workload of performing deep learning-based computationally intensive object detection on mobile edge devices is often insufficient in meeting high-precision and low-latency requirements. To address the problem of insufficient computing resources, this paper proposes a lightweight process based on a neural structure search and knowledge distillation using deep learning YOLOv8 as the baseline model. Firstly, the neural structure search algorithm was used to compress the YOLOv8 model and reduce its computational complexity. Secondly, a new knowledge distillation architecture was designed, which distills the detection head output layer and NECK feature layer to compensate for the accuracy loss caused by model reduction. When compared to YOLOv8n, the computational complexity of the lightweight model optimized in this study (in terms of floating point operations (FLOPs)) was 7.4 Gflops, which indicated a reduction of 1.3 Gflops. The multiply–accumulate operations (MACs) stood at 2.72 G, thereby illustrating a decrease of 32%; this saw an increase in the AP50, AP75, and mAP by 2.0%, 3.0%, and 1.9%, respectively. Finally, this paper designed an edge computing service architecture, and it deployed the model on the Jetson Xavier NX platform through TensorRT.

Efficient Eye State Detection for Driver Fatigue Monitoring Using Optimized YOLOv7-Tiny

This study refines the YOLOv7-tiny model through structured pruning and architectural fine-tuning, specifically for real-time eye state detection. By focusing on enhancing the model’s efficiency, particularly in environments with limited computational resources, this research contributes significantly to advancing driver monitoring systems, where timely and accurate detection of eye states such as openness or closure can prevent accidents caused by drowsiness or inattention. Structured pruning was utilized to simplify the YOLOv7-tiny model, reducing complexity and storage requirements. Subsequent fine-tuning involved adjustments to the model’s width and depth to further enhance processing speed and efficiency. The experimental outcomes reveal a pronounced reduction in storage size, of approximately 97%, accompanied by a sixfold increase in frames per second (FPS). Despite these substantial modifications, the model sustains high levels of precision, recall, and mean average precision (mAP). These improvements indicate a significant enhancement in both the speed and efficiency of the model, rendering it highly suitable for real-time applications where computational resources are limited.

A deep learning optimized LQR method for enhanced formation control with embedded systems

To achieve higher accuracy throughout the formation control processes and enhance precision in dynamic environments, particularly for the formation control of follower vehicles with embedded systems, this paper proposes a method and framework for vehicle formation control. An Ackermann-model based Linear Quadratic Regulator (LQR) controller is developed for lateral distance control and a Proportional-Derivative (PD) controller for longitudinal distance control. To enhance the efficacy of the LQR controller, the Deep Deterministic Policy Gradient Derivative (DDPG) method is introduced into the control system. The DDPG networks are trained in a simulation environment and can subsequently predict LQR parameters in real-time experiments. The practical application of the method is showcased, and the concluding remarks emphasize the potential and superior performance of our proposed formation control approach by experimental comparison with other controllers. This method can be implemented in small vehicles that possess limited computational resources and is also suitable for scenarios requiring dynamic motion control with higher tracking accuracy and stability.

Laplace-Domain Crosstalk-Free Source-Encoded Elastic Full Waveform Inversion Using Time-Domain Solvers

Crosstalk-free source-encoded elastic Full-Waveform Inversion (FWI) using time-domain solvers has demonstrated skill and efficiency at conducting seismic inversions involving multiple sources and receivers with limited computational resources. A drawback of common formulations of the procedure is that, by sweeping through the frequency domain randomly at a rate of one or a few sparsely sampled frequencies per shot, it is difficult to simultaneously incorporate time-selective data windows, as necessary for the targeting of arrivals or wave packets during the various stages of the inversion. Here, we solve this problem by using the Laplace transform of the data. Using complex-valued frequencies allows for damping the records with flexible decay rates and temporal offsets that target specific traveltimes. We present the theory of crosstalk-free source-encoded FWI in the Laplace domain, develop the details of its implementation, and illustrate the procedure with numerical examples relevant to exploration-scale scenarios.

Performance Analysis of Deep Learning Model-Compression Techniques for Audio Classification on Edge Devices

Audio classification using deep learning models, which is essential for applications like voice assistants and music analysis, faces challenges when deployed on edge devices due to their limited computational resources and memory. Achieving a balance between performance, efficiency, and accuracy is a significant obstacle to optimizing these models for such constrained environments. In this investigation, we evaluate diverse deep learning architectures, including Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM), for audio classification tasks on the ESC 50, UrbanSound8k, and Audio Set datasets. Our empirical findings indicate that Mel spectrograms outperform raw audio data, attributing this enhancement to their synergistic alignment with advanced image classification algorithms and their congruence with human auditory perception. To address the constraints of model size, we apply model-compression techniques, notably magnitude pruning, Taylor pruning, and 8-bit quantization. The research demonstrates that a hybrid pruned model achieves a commendable accuracy rate of 89 percent, which, although marginally lower than the 92 percent accuracy of the uncompressed CNN, strikingly illustrates an equilibrium between efficiency and performance. Subsequently, we deploy the optimized model on the Raspberry Pi 4 and NVIDIA Jetson Nano platforms for audio classification tasks. These findings highlight the significant potential of model-compression strategies in enabling effective deep learning applications on resource-limited devices, with minimal compromise on accuracy.

Transfer learning applications for autoencoder-based anomaly detection in wind turbines

Anomaly detection in wind turbines typically involves using normal behaviour models to detect faults early. Normal behaviour models are often implemented through the use of neural networks, of which autoencoders are particularly popular in this field. However, training autoencoder models for each turbine is time-consuming and resource intensive. Thus, transfer learning becomes essential for wind turbines with limited data or applications with limited computational resources. This study examines how cross-turbine transfer learning can be applied to autoencoder-based anomaly detection. Here, autoencoders are combined with constant thresholds for the reconstruction error to determine if input data contains an anomaly. The models are initially trained on one year’s worth of data from one or more source wind turbines. They are then fine-tuned using small amounts of data from the target wind turbine. Three methods for fine-tuning are investigated: adjusting the entire autoencoder, only the decoder, or only the threshold of the model. The performance of the transfer learning models is compared to baseline models that were trained on one year’s worth of data from the target wind turbine. The results of the tests conducted in this study indicate that models trained on data of multiple wind turbines do not improve the anomaly detection capability compared to models trained on data of one source wind turbine. In addition, modifying the model’s threshold can lead to comparable or even superior performance compared to the baseline, whereas fine-tuning the decoder or autoencoder further enhances the models’ performance.

Maximizing throughput in NOMA-enable industrial IoT networks using digital twin and reinforcement learning

IntroductionIncreased deployment of heterogeneous and complex Industrial Internet of Things (IIoT) applications such as predictive maintenance and asset tracking places a substantial strain on the limited computational and communication resources. To cater to the rigorous demands of these applications, it is imperative to devise an adaptive online resource allocation method to enhance the efficiency of the current network operations. Multiaccess edge computing (MEC) and digital twins (DTs) are promising solutions that facilitate the realization of edge intelligence and find applications in various industrial applications. Yet, little is known about the advantage the two technologies offer to IIoT networks. ObjectiveThis study presents a joint optimization of offloading and resource allocation approach where MEC-server DT is created at the edge, and nonorthogonal multiple access (NOMA) communication is considered between IIoT devices and the industrial gateways (IGWs) for spectral efficiency. Our proposed framework is tailored to reduce mean task completion latency and enhance overall IIoT network throughput. MethodTo achieve our objective, we jointly optimize the computation resource allocation (RA), subchannel assignment (SA), and offloading decisions (OD). Given the inherent complexity of the problem, we further divide it into RA and SA/OD sub-problems. Employing Deep Reinforcement Learning (DRL), we have formulated a solution delineating the most efficient RA strategy and leveraged DT for optimal SA/OD strategies. ResultsSimulation results demonstrate the superior efficiency of our framework, realizing up to 92 % of the efficiency of the exhaustive search method while reducing computation and action decision time. ConclusionIn light of system dynamics considered for our work, the proposed framework perfomance showcase its robustness and potential application in real-world IIoT networks.

Towards Efficient Coordination in Multi-Agent Reinforcement Learning through Hybrid Information-Driven Approaches

Abstract: This paper introduces a novel framework to enhance coordination in multi-agent reinforcement learning (MARL) systems by integrating mutual information (MMI) principles with information-driven strategies. Firstly, we propose a variational approach leveraging MMI to promote coordinated behaviors among agents by regulating the cumulative return alongside the simultaneous mutual information between multi-agent actions. Through the introduction of a latent variable inducing nonzero mutual information and the application of a variational bound, a tractable lower bound is derived for the MMI-regularized objective function. This bound combines maximum entropy reinforcement learning with reducing uncertainty in other agents' actions. Subsequently, we present a practical algorithm, Variational Maximum Mutual Information Multi-Agent Actor-Critic (VM3-AC), utilizing policy iteration to maximize the derived lower bound, following a centralized learning with decentralized execution (CTDE) paradigm. Secondly, we explore the challenges of large state spaces and limited computational resources in distributed multi-agent systems, proposing a hybrid information-driven MARL approach. This approach integrates informationtheoretic models as heuristics to aid navigation in sparse state spaces, complemented by information-based rewards within an RL framework to learn higher-level policies efficiently. Our preliminary findings suggest that this hybrid approach could enhance exploration efficiency significantly, demonstrating approximately three orders of magnitude improvement over naive baseline metrics. Although still in its early stages, this work presents a promising direction for future research in achieving efficient coordination in MARL systems