AAAnkit Agarwal

Ankit Agarwal

*This is an algorithmically generated author record based on publicly available information.

Ankit Agarwal has a total of 16 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Enhancing Mobility and Rehabilitation

    by Ankit Agarwal, Gom Taye, Indra Kishor, V. Akilandeswari

    This study aims to develop and evaluate an intelligent wearable-exoskeleton system designed to assist and enhance lower limb rehabilitation in individuals with mobility impairments. The proposed system combines inertial measurement units (IMUs), surface electromyography. A hybrid deep learning model (CNN-LSTM) is deployed on an edge computing unit for real-time movement classification. Data from 10 participants with mild-to-moderate lower limb impairments were collected during controlled rehabilitation tasks, including walking, stair climbing, and sit-to-stand transitions. Quantitative performance metrics such as gait velocity, step symmetry, and muscle activation were analyzed pre- and post-intervention. Usability was assessed using the System Usability Scale (SUS), while response latency and classification accuracy were logged. Post-intervention results demonstrated a 23.9% increase in gait velocity and a 15.2% improvement in step symmetry. Muscle effort, indicated by RMS sEMG values, decreased by 12.4%. The CNN-LSTM model achieved 94.2% accuracy with <10 ms inference latency.


  • Virtual Reality (VR) and Augmented Reality (AR) Simulators for Medical Education

    by Indra Kishor, Manvi Sharma, Anusha Preetham, Ankit Agarwal, Udit Mamodiya

    Medical education has shift with the immersive technologies like VR and AR that provide realistic and interactive learning. Although current VR/AR simulators have improved surgical training, anatomy visualization, and clinical decision-making, they do not provide real-time adaptive learning and multi-sensory feedback. A novel AI-enhanced VR/AR medical simulator combining real-time performance analysis, haptic feedback, and cloud collaboration to provide a learning environment for interactive and adaptive learning for medical students. It also provides haptic feedback integration with realistic surgical precision and procedure training. A prototype system of this type is implemented with emphasis on surgical simulation and emergency response training. Experimental evaluation of the system tests learning efficiency, error rate reduction, cognitive engagement, and haptic response accuracy against traditional learning techniques. Early findings show that there is considerable gain in skill retention and efficiency of training by employing the AI-based VR/AR simulator methods.


  • Leveraging Digital Twin Technology in Industrial IoT for Energy Optimization and Waste Reduction

    by Ankit Agarwal, Inderjeet Sinha, Subarno Bhattacharyya, Udit Mamodiya

    It was based on the motivation to generate novel innovative technologies in order to promote energy efficiency with reduced waste through the growth in demand for industrial practices in an eco-friendly way. In this paradigm, Industrial IoT provides a promising platform for real-time monitoring, maintenance, and resource usage within Digital Twins. A new multi-layer architecture will be developed in this chapter to integrate digital twins and IIoT to mutually optimize energy consumption and waste reduction. Named the Adaptive Energy Optimization and Waste Reduction Framework, this novel architecture utilizes AI-driven self-learning models for dynamically adapting itself to various changes in industrial conditions. Simultaneously, efficiency is maintained through both energy usage as well as waste minimization due to EWCO. This framework is simulated to carry out industrial operations in virtual environments to identify inefficiencies and predict failures along with the automation of optimization strategies.

  • Adaptive Multi-Modal Neural Network for Real-Time Threat Detection

    by Indra Kishor, Udit Mamodiya, Ankit Agarwal, Abhishek Bhattacherjee

    Real-time threat detection is crucial in domains like cybersecurity, surveillance, and public safety in general, where quick and effective detection of risks is critical. Current detection approaches suffer from limitations such as high latency, in-processing multi-modal data efficiently, and learning to adapt to evolvement of threats. Therefore, this paper proposes a new algorithm called Adaptive Multi-Modal Neural Network, or simply AMM-Net for a novel neural network algorithm for overcoming the above-listed limitations. AMM-Net incorporates the use of CNNs, RNNs, and transformers in fusing visual, sequential, and textual data streams. Its use of a dynamic attention-based fusion layer allows it to dynamically focus on the most relevant features across modalities for an efficient detection in a myriad and complex scenario. Techniques that aid in real-time deployment through the use of model pruning, quantization, and edge computing will further make it deployable to resource-constrained environments.


  • Open Access

    Ensuring Cloud Tenant Data Security and OptimizingDDoS Detection Efficiency through Trust-Based Filtering

    by Ankit Agarwal, Saurabh Gupta, G Swamy, G Prasadu, Vandana Ahuja, Ajay Kumar Badhan, Udit Mamodiya

    Tenant data security has grown critical as more and more commercial enterprises use cloud services for data processing and storage. The present study endeavors to tackle the dual problems of preserving tenant data in the cloud and enhancing the effectiveness of detecting Distributed Denial of Service (DDoS) attacks. We provide a brand-new strategy called Trust-Based Filtering that incorporates trust measurements into the current cloud security framework. Our approach for data security is centered on access controls, encryption, and ongoing monitoring. Sensitive information is prevented from being accessed by unauthorized parties even in the case of a breach thanks to encrypted data storage. Based on trust scores, access controls are dynamically modified to grant genuine tenants more access privileges while limiting dubious entities. The dynamic trust model used by the Trust-Based Filtering system assesses cloud tenants' behavior by looking at their past interactions with the cloud environment.