Praveen Kumar
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Praveen Kumar has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
Durai Arumugam S S L
1 shared publication- Durai Arumugam Sivakolunthu Sreevelu Latha
Department of Information Technology, Easwari Engineering College, Ramapuram, Chennai, Tamil Nadu 600089, India
1 shared publication Jennifer S. Raj
Department of Electronics and Communication Engineering Gnanamani College of Technology Namakkal India
1 shared publication- Krishnamohan Reddy Kunduru
Department of Engineering Design, Overhead Door Corporation, Lewisville, USA
1 shared publication - Nanda Harishankar Krishna
Universite de Montreal, Mila - Quebec Artificial Intelligence Institute
1 shared publication - R. Muthukumar
Department of Electrical and Electronics Engineering, Erode Sengunthar Engineering College, India
1 shared publication
Latest Publications
- 28 June 2024
- Published by World Scientific Pub Co Pte Ltd in International Journal on Artificial Intelligence Tools
- Vol. 33 (04)
Hybrid Optimized Gated Recurrent Unit with Ridge Classifier for Crop Recommendation for Precise Agriculture Using Fused Feature Selection Concept
Agriculture is considered the leading field around the world, which is also the backbone of India. Agriculture is in a flawed state because the temperature changes, along with their uncertainty, cause huge damage to the crops during the manufacturing process. So, the appropriate prediction of crop expansion plays a vital role in the management of crop growth. This prediction can enhance the federated industries to make their sustainability toward the occupation. Recently, the farmers have not selected suitable crops for their cultivation based on soil factors. This makes a negative impact on crop yield, and thus, the Indian farmers can suffer from severe losses besides the monetary front. Hence, the optimal crop recommendation model has to consider different parameters of the soil for forecasting the best crop for cultivation, which increases crop growth and crop production. Thus, this research work explores a new crop recommendation model for precision agriculture intending to promote crop yield and alleviate the loss to farmers. Initially, this research work gathers the standard data regarding the agricultural parameters of some areas. Then, the deep features using an autoencoder, and statistical features are gathered along with the Principal Component Analysis (PCA)-based features. Next, all three sets of features are fused and fed to the developed Adaptive Henry Gas Solubility Optimization (AHGSO) for selecting the optimal features. Finally, the chosen optimal features are fed to the recommendation stage, where a Gated Recurrent Unit with Ridge Classifier (GRU-RC) is suggested for getting the precise outcome regarding the recommended crop suitable to that agricultural parameter. Here, the optimal solutions are attained by tuning the parameters of GRU and ridge classifier with the same I-HGSO. At last, the results obtained from the hybrid method can be considered more efficient.
- 28 June 2024
- Published by World Scientific Pub Co Pte Ltd in International Journal on Artificial Intelligence Tools
- Vol. 33 (04)
Hybrid Optimized Gated Recurrent Unit with Ridge Classifier for Crop Recommendation for Precise Agriculture Using Fused Feature Selection Concept
Agriculture is considered the leading field around the world, which is also the backbone of India. Agriculture is in a flawed state because the temperature changes, along with their uncertainty, cause huge damage to the crops during the manufacturing process. So, the appropriate prediction of crop expansion plays a vital role in the management of crop growth. This prediction can enhance the federated industries to make their sustainability toward the occupation. Recently, the farmers have not selected suitable crops for their cultivation based on soil factors. This makes a negative impact on crop yield, and thus, the Indian farmers can suffer from severe losses besides the monetary front. Hence, the optimal crop recommendation model has to consider different parameters of the soil for forecasting the best crop for cultivation, which increases crop growth and crop production. Thus, this research work explores a new crop recommendation model for precision agriculture intending to promote crop yield and alleviate the loss to farmers. Initially, this research work gathers the standard data regarding the agricultural parameters of some areas. Then, the deep features using an autoencoder, and statistical features are gathered along with the Principal Component Analysis (PCA)-based features. Next, all three sets of features are fused and fed to the developed Adaptive Henry Gas Solubility Optimization (AHGSO) for selecting the optimal features. Finally, the chosen optimal features are fed to the recommendation stage, where a Gated Recurrent Unit with Ridge Classifier (GRU-RC) is suggested for getting the precise outcome regarding the recommended crop suitable to that agricultural parameter. Here, the optimal solutions are attained by tuning the parameters of GRU and ridge classifier with the same I-HGSO. At last, the results obtained from the hybrid method can be considered more efficient.
- 21 June 2024
- Published by IGI Global Scientific Publishing
- p. 312-337
Leveraging Fuel Cell Technology With AI and ML Integration for Next-Generation Vehicles
The integration of fuel cell technology with artificial intelligence and machine learning in electric vehicles (EVs) has the potential to enhance efficiency, performance, and reliability. Fuel cells, a clean alternative to traditional engines, produce electricity through electrochemical reactions between hydrogen and oxygen, with water vapor as the only byproduct. AI-driven algorithms analyze vast data from sensors and onboard systems, while ML algorithms enable early detection of potential system failures. AI-based driver assist systems can optimize driving behaviors using fuel cell data. However, integrating fuel cell technology with AI and ML faces challenges like data management, algorithm development, and interoperability with existing vehicle systems. Interdisciplinary collaboration between automotive engineers, data scientists, and AI specialists is needed to develop robust, scalable solutions.
- 21 June 2024
- Published by IGI Global Scientific Publishing
- p. 312-337
Leveraging Fuel Cell Technology With AI and ML Integration for Next-Generation Vehicles
The integration of fuel cell technology with artificial intelligence and machine learning in electric vehicles (EVs) has the potential to enhance efficiency, performance, and reliability. Fuel cells, a clean alternative to traditional engines, produce electricity through electrochemical reactions between hydrogen and oxygen, with water vapor as the only byproduct. AI-driven algorithms analyze vast data from sensors and onboard systems, while ML algorithms enable early detection of potential system failures. AI-based driver assist systems can optimize driving behaviors using fuel cell data. However, integrating fuel cell technology with AI and ML faces challenges like data management, algorithm development, and interoperability with existing vehicle systems. Interdisciplinary collaboration between automotive engineers, data scientists, and AI specialists is needed to develop robust, scalable solutions.
- 23 May 2024
- Published by Elsevier in Applied Soft Computing
- Vol. 162
Serial cascaded deep feature extraction-based adaptive attention dilated model for crop recommendation framework
- 23 May 2024
- Published by Elsevier in Applied Soft Computing
- Vol. 162
Serial cascaded deep feature extraction-based adaptive attention dilated model for crop recommendation framework
- 14 September 2021
- Published by Wiley in International Journal of Communication Systems
- Vol. 34 (17) , e4978
Analysis of dynamic topology wireless sensor networks for the internet of things (IoT)
Summary: The availability of the smart existing systems utilizes the system assets for crucial internet of things (IoT). The IoT visions to develop effectively with the processing model. Past customary versatile figuring set‐ups utilize advanced mobile phones and portables. A sensor could drop out for various reasons, for the system to act naturally adjusting and consider multipath steering. Dynamic topology for a remote sensor arranges all the detecting hubs that are associated together for gathering information that are transmitted. The goals that are accompanying are set for exploring the work that are to be finished. Several sensor hubs in the system are arranged for the improvement of proficiency dynamic topology. The novel strategy has been proposed for the improvement in oneself sorting out, multi‐bounce systems for the remote sensor hubs. Because of its ability to decrease energy utilisation, a cluster‐based model is the best in the remote sensor arrangement. The cluster head selection processes are a lumbering procedure that influences execution on the system execution. Few investigations propose the determination of the cluster head techniques that are being large portion of them that are not fitting for a powerful grouping condition. The calculations for the proposed methodology work better inside a solitary bounce grouping model structure, and the system lifetime establishes a major issue if there should be an occurrence of multi‐jump bunching conditions. The manuscript presents cluster head determination technique novel single‐hop and multi‐hop cluster head selection by GA (S/M HCH‐GA) in wireless sensor networks (WSNs) for both one‐hop as well as multi‐hop cluster models. The technique that is proposed intends to meet the prerequisites of dynamic situations; this is done by choosing cluster head that is dependent on the energy that remains along with the energy that is consumed.
- 14 September 2021
- Published by Wiley in International Journal of Communication Systems
- Vol. 34 (17) , e4978
Analysis of dynamic topology wireless sensor networks for the internet of things (IoT)
Summary: The availability of the smart existing systems utilizes the system assets for crucial internet of things (IoT). The IoT visions to develop effectively with the processing model. Past customary versatile figuring set‐ups utilize advanced mobile phones and portables. A sensor could drop out for various reasons, for the system to act naturally adjusting and consider multipath steering. Dynamic topology for a remote sensor arranges all the detecting hubs that are associated together for gathering information that are transmitted. The goals that are accompanying are set for exploring the work that are to be finished. Several sensor hubs in the system are arranged for the improvement of proficiency dynamic topology. The novel strategy has been proposed for the improvement in oneself sorting out, multi‐bounce systems for the remote sensor hubs. Because of its ability to decrease energy utilisation, a cluster‐based model is the best in the remote sensor arrangement. The cluster head selection processes are a lumbering procedure that influences execution on the system execution. Few investigations propose the determination of the cluster head techniques that are being large portion of them that are not fitting for a powerful grouping condition. The calculations for the proposed methodology work better inside a solitary bounce grouping model structure, and the system lifetime establishes a major issue if there should be an occurrence of multi‐jump bunching conditions. The manuscript presents cluster head determination technique novel single‐hop and multi‐hop cluster head selection by GA (S/M HCH‐GA) in wireless sensor networks (WSNs) for both one‐hop as well as multi‐hop cluster models. The technique that is proposed intends to meet the prerequisites of dynamic situations; this is done by choosing cluster head that is dependent on the energy that remains along with the energy that is consumed.
- 1 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1281-1288
Defending Against Localized Adversarial Attacks on Edge-Deployed Monocular Depth Estimators
Estimation of depth from a single image is an important scene understanding task in computer vision. With the advent of Deep Learning and Convolutional Neural Networks, staggeringly high accuracies have been achieved in this task. With advancements in model optimization, it has been possible to deploy these models on edge devices, allowing for efficient depth estimation in safety-critical applications in robots, rovers, drones and even self-driving vehicles. However, these models are susceptible to attacks from malicious adversaries, which aim to distort the output of the model for a seemingly clean image by adding minute perturbations. In the real-world scenario, the most plausible attack is the adversarial patch, which can be printed and used as a physical adversarial attack against Deep Learning models. In the case of Monocular Depth Estimation, we show that small adversarial patches, which range from 0.7% to 5% of the image size, greatly worsen model performance. It is thus essential that these models are made robust using defense mechanisms, to defend against malicious inputs while also not reducing performance on clean images. Moreover, it is essential that the defense mechanism be computationally efficient, for real-time inference on edge devices. In this work, we propose the first defense mechanism against adversarial patches for a regression network, in the context of Monocular Depth Estimation on an edge device. The defense mechanism adds very little overhead time of 38 milliseconds on a Raspberry Pi 3 Model B, maintaining performance on clean images while also achieving near clean image levels of performance on adversarial inputs.
- 1 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1281-1288
Defending Against Localized Adversarial Attacks on Edge-Deployed Monocular Depth Estimators
Estimation of depth from a single image is an important scene understanding task in computer vision. With the advent of Deep Learning and Convolutional Neural Networks, staggeringly high accuracies have been achieved in this task. With advancements in model optimization, it has been possible to deploy these models on edge devices, allowing for efficient depth estimation in safety-critical applications in robots, rovers, drones and even self-driving vehicles. However, these models are susceptible to attacks from malicious adversaries, which aim to distort the output of the model for a seemingly clean image by adding minute perturbations. In the real-world scenario, the most plausible attack is the adversarial patch, which can be printed and used as a physical adversarial attack against Deep Learning models. In the case of Monocular Depth Estimation, we show that small adversarial patches, which range from 0.7% to 5% of the image size, greatly worsen model performance. It is thus essential that these models are made robust using defense mechanisms, to defend against malicious inputs while also not reducing performance on clean images. Moreover, it is essential that the defense mechanism be computationally efficient, for real-time inference on edge devices. In this work, we propose the first defense mechanism against adversarial patches for a regression network, in the context of Monocular Depth Estimation on an edge device. The defense mechanism adds very little overhead time of 38 milliseconds on a Raspberry Pi 3 Model B, maintaining performance on clean images while also achieving near clean image levels of performance on adversarial inputs.