Malak Abid Ali Khan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Malak Abid Ali Khan has a total of 57 co-authors. Here are the top co-authors by number of shared publications:
- Hongbin Ma
School of Automation, Beijing Institute of Technology, Beijing 100081, China
8 shared publications Hongbin Ma
National Key Lab of Autonomous Intelligent Unmanned Systems, School of Automation Beijing Institute of Technology Beijing China
6 shared publicationsZia Ur Rehman
5 shared publicationsImran Khan Mirani
4 shared publications- Senlin Luo
School of Information and Electronics, Beijing Institute of Technology, Beijing 100081, China
4 shared publications Syed Muhammad Aamir
4 shared publications
Latest Publications
- 29 September 2026
- Published by Elsevier in Computer Networks
- Vol. 290
LoRaWAN-based federated learning for the internet of vehicles: a comprehensive survey of architectures, algorithms, and open challenges
- 29 September 2026
- Published by Elsevier in Computer Networks
- Vol. 290
LoRaWAN-based federated learning for the internet of vehicles: a comprehensive survey of architectures, algorithms, and open challenges
- 4 September 2026
- Published by AIP Publishing in The Journal of Chemical Physics
- Vol. 165 (9)
Lithium solvation and interfacial structure in a ternary carbonate electrolyte at electrified graphite interface: Constant charge vs constant potential molecular dynamics
Multicomponent carbonate electrolytes are central to next generation electrochemical energy storage devices, and the choice of electrode model strongly shapes their predicted interfacial behavior. This study compares the constant charge method (CCM) and constant potential method (CPM) for modeling electrified interfaces in such electrolytes. Molecular dynamic simulations are performed for 1 M LiPF6 in an ethylene carbonate:dimethyl carbonate:ethyl methyl carbonate ternary carbonate electrolyte sandwiched between two graphite electrodes. Results show that CCM and CPM give nearly identical interfacial structures at low and moderate polarization, with clear deviations emerging at ±1.35 V, equivalent to a 2.70 V cell voltage. This method-dependent response originates mainly from Li+ behavior at the negative electrode, where CCM promotes partial desolvation and contact-like adsorption, whereas CPM preserves solvent separated adsorption through dynamic charge redistribution and stronger interfacial solvent ordering. Meanwhile, PF6− distributions remain comparatively insensitive to electrode treatment. This study defines the polarization regime where CCM is sufficient and establishes when CPM is required to capture coupled Li+/solvent/electrode interactions in carbonate solvent-based energy storage electrolytes.
- 4 September 2026
- Published by AIP Publishing in The Journal of Chemical Physics
- Vol. 165 (9)
Lithium solvation and interfacial structure in a ternary carbonate electrolyte at electrified graphite interface: Constant charge vs constant potential molecular dynamics
Multicomponent carbonate electrolytes are central to next generation electrochemical energy storage devices, and the choice of electrode model strongly shapes their predicted interfacial behavior. This study compares the constant charge method (CCM) and constant potential method (CPM) for modeling electrified interfaces in such electrolytes. Molecular dynamic simulations are performed for 1 M LiPF6 in an ethylene carbonate:dimethyl carbonate:ethyl methyl carbonate ternary carbonate electrolyte sandwiched between two graphite electrodes. Results show that CCM and CPM give nearly identical interfacial structures at low and moderate polarization, with clear deviations emerging at ±1.35 V, equivalent to a 2.70 V cell voltage. This method-dependent response originates mainly from Li+ behavior at the negative electrode, where CCM promotes partial desolvation and contact-like adsorption, whereas CPM preserves solvent separated adsorption through dynamic charge redistribution and stronger interfacial solvent ordering. Meanwhile, PF6− distributions remain comparatively insensitive to electrode treatment. This study defines the polarization regime where CCM is sufficient and establishes when CPM is required to capture coupled Li+/solvent/electrode interactions in carbonate solvent-based energy storage electrolytes.
- 5 July 2026
- Published by Wiley in Quality and Reliability Engineering International
- Vol. 42 (7) , 4113-4133
FAD‐Net: An Efficient Architecture for Defect Recognition in Semiconductor Wafers
Automatic wafer map classification and recognition remain a major challenge in the intelligent semiconductor manufacturing process. The defect category identified from a wafer map provides valuable information for tracing the primary cause to improve production quality through classification and recognition of wafer map defect patterns. In this work, a novel architecture is proposed, termed the Feature Aware Dendritic Network (FAD‐Net), for pattern recognition and multi‐class classification of complex single‐type wafer map defects. It integrates a Dendritic Neuron Module (DNM), a Feature Pyramid Module (FPM), and attention mechanisms to capture discriminative features. The three‐stage model uses a biologically inspired DNM to focus on nonlinear interactions among features, an FPM for aggregating multi‐scale features, and attention mechanisms for enhancing spatial and channel‐wise feature representations to improve the classification and recognition. Experimental results demonstrate an accuracy of 95.35%–98.63%, F1‐score of 0.9255–0.9858, and an AUC of 0.9971–0.9997 on WM‐811K and MixedWM38 datasets. Compared to state‐of‐the‐art, FAD‐Net delivers improved recognition and classification performance for wafer map defect analysis.
- 5 July 2026
- Published by Wiley in Quality and Reliability Engineering International
- Vol. 42 (7) , 4113-4133
FAD‐Net: An Efficient Architecture for Defect Recognition in Semiconductor Wafers
Automatic wafer map classification and recognition remain a major challenge in the intelligent semiconductor manufacturing process. The defect category identified from a wafer map provides valuable information for tracing the primary cause to improve production quality through classification and recognition of wafer map defect patterns. In this work, a novel architecture is proposed, termed the Feature Aware Dendritic Network (FAD‐Net), for pattern recognition and multi‐class classification of complex single‐type wafer map defects. It integrates a Dendritic Neuron Module (DNM), a Feature Pyramid Module (FPM), and attention mechanisms to capture discriminative features. The three‐stage model uses a biologically inspired DNM to focus on nonlinear interactions among features, an FPM for aggregating multi‐scale features, and attention mechanisms for enhancing spatial and channel‐wise feature representations to improve the classification and recognition. Experimental results demonstrate an accuracy of 95.35%–98.63%, F1‐score of 0.9255–0.9858, and an AUC of 0.9971–0.9997 on WM‐811K and MixedWM38 datasets. Compared to state‐of‐the‐art, FAD‐Net delivers improved recognition and classification performance for wafer map defect analysis.
- Open Access
- 31 March 2026
- Published by Elsevier in SSRN Electronic Journal
ADEO: An Anomaly Detection and Energy Optimization for Centralized LoRa-IoT Networks based on Spreading Factor
In Centralized LoRaWAN networks, frequent retransmissions due to attacks severely degrade end device battery life. Existing methods overlook criti cal LoRa-specific parameters such as Spreading Factor (SF) and Received Signal Strength Indicator (RSSI), and lack unified energy-efficient detection mitigation frameworks. We propose ADEO, an Anomaly Detection and Energy Optimization method that integrates an Autoencoder (AE) with an enhanced Gaussian Mixture Model (GMM). RSSI sequences are labeled per optimal SF and used to train a hybrid detector where AE captures reconstruc tion errors and GMM is improved via mutual assistance of latent variables and Dynamic Latent Space Weighting (DLSW) models’ probabilistic likeli hood. Their outputs are fused via majority voting, achieving more detection over the individual models. Upon anomaly detection, energy is saved via adaptive SF downgrading and retransmission blocking under poor channel conditions. Evaluated on three real-world and one synthetic LoRa dataset, ADEOachieves 8.4%–31.7% higher detection rates while maintaining energy efficiency, making it suitable for resource-constrained IoT deployments.
- 31 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessADEO: An Anomaly Detection and Energy Optimization for Centralized LoRa-IoT Networks based on Spreading Factor
In Centralized LoRaWAN networks, frequent retransmissions due to attacks severely degrade end device battery life. Existing methods overlook criti cal LoRa-specific parameters such as Spreading Factor (SF) and Received Signal Strength Indicator (RSSI), and lack unified energy-efficient detection mitigation frameworks. We propose ADEO, an Anomaly Detection and Energy Optimization method that integrates an Autoencoder (AE) with an enhanced Gaussian Mixture Model (GMM). RSSI sequences are labeled per optimal SF and used to train a hybrid detector where AE captures reconstruc tion errors and GMM is improved via mutual assistance of latent variables and Dynamic Latent Space Weighting (DLSW) models’ probabilistic likeli hood. Their outputs are fused via majority voting, achieving more detection over the individual models. Upon anomaly detection, energy is saved via adaptive SF downgrading and retransmission blocking under poor channel conditions. Evaluated on three real-world and one synthetic LoRa dataset, ADEOachieves 8.4%–31.7% higher detection rates while maintaining energy efficiency, making it suitable for resource-constrained IoT deployments.
- 10 January 2026
- Published by Springer Nature
- p. 116-129
Robust Road Surface Defects Detection Using Three-Dimensional Geometric Analysis and Adaptive Machine Learning Strategy
- 10 January 2026
- Published by Springer Nature
- p. 116-129
Robust Road Surface Defects Detection Using Three-Dimensional Geometric Analysis and Adaptive Machine Learning Strategy
- 7 December 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 303
Dual resource allocation for enhancing vehicle tracking using deep neural network and Bayesian LoRa
- 7 December 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 303
Dual resource allocation for enhancing vehicle tracking using deep neural network and Bayesian LoRa
- 4 September 2025
- Published by Wiley in Internet Technology Letters
- Vol. 8 (5)
Enhancing Car Tracking Systems With DNN‐LoRa
This paper outlines a fusion of deep neural network and LoRa technology for car tracking optimization. LoRa's SX1301 gateway (GW) applies the Bayesian game parameter selection (BGPS) approach for switching the transmission power at the network server. At the same time, the car node (CN) uses a hybrid model to change the spreading factor and data rate. By reducing power losses among GWs, BGPS substantially increases the packet success rate (PSR) at the CN. The hybrid model enables adaptive decision‐making, resulting in improved tracking precision and reduced latency with efficient energy usage. However, it exhibits a 95.9% PSR with increased latency noted at the lower bandwidth.
- 4 September 2025
- Published by Wiley in Internet Technology Letters
- Vol. 8 (5)
Enhancing Car Tracking Systems With DNN‐LoRa
This paper outlines a fusion of deep neural network and LoRa technology for car tracking optimization. LoRa's SX1301 gateway (GW) applies the Bayesian game parameter selection (BGPS) approach for switching the transmission power at the network server. At the same time, the car node (CN) uses a hybrid model to change the spreading factor and data rate. By reducing power losses among GWs, BGPS substantially increases the packet success rate (PSR) at the CN. The hybrid model enables adaptive decision‐making, resulting in improved tracking precision and reduced latency with efficient energy usage. However, it exhibits a 95.9% PSR with increased latency noted at the lower bandwidth.
- 30 January 2025
- Published by Wiley in Transactions on Emerging Telecommunications Technologies
- Vol. 36 (2)
LoRa Meets Artificial Intelligence to Optimize Indoor Networks for Static EDs
The architectural design of the Indoor Internet of Things (IIoT) network targeting static end devices (EDs) and gateways (GWs) has been innovatively formulated in this paper, integrating LoRa technology to mitigate losses and ensure seamless information reception through meticulous ED allocation. The arrangement of simultaneously transmitted data within the network server (NS) employs a deep neural network (DNN) with distributed machine learning (DML) to adjust transmission parameters, ensuring frequent uninterrupted bidirectional communication. This augmentation is obtained by strategically deploying EDs within distinct clusters determined by K‐means and density‐based spatial clustering with noise (DBSCAN), thus optimizing spreading factor (SF) and data rate (DR) allocation to prevent data congestion and improve signal‐to‐interference noise ratio (SINR). The proposed hybrid model (DR|SF) for pure and slotted ALOHA amplifies the network's performance metrics for indoor scenarios. A unified framework utilizing a one‐slope model estimates path losses (PL) while exploring various bandwidths (BW), bidirectional interrogations, and duty cycles (DC) to lower the saturation and prolong the active lifespan of the EDs. The results manifest a packet rejection rate (PRR) of 0% for the DBSCAN, contrasting a 4.7% estimate for the K‐means. The network saturation is minimized to 9.5% and 10.1%, correspondingly, significantly increasing the efficiency of slotted ALOHA (91%) and pure ALOHA (90.6%), thereby prolonging the longevity of EDs.
- 30 January 2025
- Published by Wiley in Transactions on Emerging Telecommunications Technologies
- Vol. 36 (2)
LoRa Meets Artificial Intelligence to Optimize Indoor Networks for Static EDs
The architectural design of the Indoor Internet of Things (IIoT) network targeting static end devices (EDs) and gateways (GWs) has been innovatively formulated in this paper, integrating LoRa technology to mitigate losses and ensure seamless information reception through meticulous ED allocation. The arrangement of simultaneously transmitted data within the network server (NS) employs a deep neural network (DNN) with distributed machine learning (DML) to adjust transmission parameters, ensuring frequent uninterrupted bidirectional communication. This augmentation is obtained by strategically deploying EDs within distinct clusters determined by K‐means and density‐based spatial clustering with noise (DBSCAN), thus optimizing spreading factor (SF) and data rate (DR) allocation to prevent data congestion and improve signal‐to‐interference noise ratio (SINR). The proposed hybrid model (DR|SF) for pure and slotted ALOHA amplifies the network's performance metrics for indoor scenarios. A unified framework utilizing a one‐slope model estimates path losses (PL) while exploring various bandwidths (BW), bidirectional interrogations, and duty cycles (DC) to lower the saturation and prolong the active lifespan of the EDs. The results manifest a packet rejection rate (PRR) of 0% for the DBSCAN, contrasting a 4.7% estimate for the K‐means. The network saturation is minimized to 9.5% and 10.1%, correspondingly, significantly increasing the efficiency of slotted ALOHA (91%) and pure ALOHA (90.6%), thereby prolonging the longevity of EDs.
- 28 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 137-142
Optimization of LoRa for Distributed Environments Based on Machine Learning
In the rapidly evolving Internet of Things (IoT) field, wireless communication technologies have revolutionized industries by connecting smart devices for extensive data sharing and automation. However, distributed wireless communication systems often face limited signal coverage and high maintenance costs. This paper introduces optimization techniques based on machine learning in a distributed environment, aiming to design a low-power and long-range LoRa network for indoor and outdoor IoT applications. The orthogonal combinations of transmission parameters and the K-means approach effectively address the avalanche effects and maximize the throughput of its data collision avoidance algorithm (ALOHA protocol) for improving the network's overall performance.
- 28 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 137-142
Optimization of LoRa for Distributed Environments Based on Machine Learning
In the rapidly evolving Internet of Things (IoT) field, wireless communication technologies have revolutionized industries by connecting smart devices for extensive data sharing and automation. However, distributed wireless communication systems often face limited signal coverage and high maintenance costs. This paper introduces optimization techniques based on machine learning in a distributed environment, aiming to design a low-power and long-range LoRa network for indoor and outdoor IoT applications. The orthogonal combinations of transmission parameters and the K-means approach effectively address the avalanche effects and maximize the throughput of its data collision avoidance algorithm (ALOHA protocol) for improving the network's overall performance.