Shahnewaz Siddique

Md. Rafat Rahman Tushar, Shahnewaz Siddique

To perform well, Deep Reinforcement Learning (DRL) methods require significant memory resources and computational time. Also, sometimes these systems need additional environment information to achieve a good reward. However, it is more important for many applications and devices to reduce memory usage and computational times than to achieve the maximum reward. This paper presents a modified DRL method that performs reasonably well with compressed imagery data without requiring additional environment information and also uses less memory and time. We have designed a lightweight Convolutional Neural Network (CNN) with a variant of the Q-network that efficiently takes preprocessed image data as input and uses less memory. Furthermore, we use a simple reward mechanism and small experience replay memory so as to provide only the minimum necessary information. Our modified DRL method enables our autonomous agent to play Snake, a classical control game. The results show our model can achieve similar performance as other DRL methods.

Ahmed Faizul Haque Dhrubo, Souvik Pramanik, Most. Aysha Siddika Sumona et al.

This study introduces a novel architecture of KAN-based BiGRU model for the task of classification and summarization of legal documents in a low-resource multilingual setup. In order to tackle problems associated with domain language, the usage of different languages, long dependencies within context, and class imbalance, we employ the dataset composed of legal documents from Bangladesh and taken from Manupatra, which include Bengali, English, and transliterated Bengali languages. Our classification task involves BiGRU model, along with Kolmogorov-Arnold Network (KAN) module, while the summarization part utilizes attention-based GRU, combined with a KAN model head. Classification model yields 67.96% of accuracy and 0.65 F1 score; while ROUGE-1, ROUGE-2, and ROUGE-L measures for summarization yield 0.38, 0.23, and 0.31 F1 scores, correspondingly. Ablation study shows that the use of KAN increases classification accuracy from 57.34% to 67.96%. Moreover, our proposed technique is compared to several baselines, including classical ML algorithms and pretrained language models.

Saikat Barua, Mostafizur Rahman, Shehenaz Khaled et al.

The emergence of hybrid quantum-classical machine learning (HQML) models opens new horizons of computational intelligence but their fundamental complexity frequently leads to black box behavior that undermines transparency and reliability in their application. Although XAI for quantum systems still in its infancy, a major research gap is evident in robust global and local explainability approaches that are designed for HQML architectures that employ quantized feature encoding followed by classical learning. The gap is the focus of this work, which introduces QuXAI, an framework based upon Q-MEDLEY, an explainer for explaining feature importance in these hybrid systems. Our model entails the creation of HQML models incorporating quantum feature maps, the use of Q-MEDLEY, which combines feature based inferences, preserving the quantum transformation stage and visualizing the resulting attributions. Our result shows that Q-MEDLEY delineates influential classical aspects in HQML models, as well as separates their noise, and competes well against established XAI techniques in classical validation settings. Ablation studies more significantly expose the virtues of the composite structure used in Q-MEDLEY. The implications of this work are critically important, as it provides a route to improve the interpretability and reliability of HQML models, thus promoting greater confidence and being able to engage in safer and more responsible use of quantum-enhanced AI technology. Our code and experiments are open-sourced at: https://github.com/GitsSaikat/QuXAI

Rotan Hawlader Pranto, Shahnewaz Siddique

The human body communicates through various meaningful gestures, with sign language using hands being a prominent example. Bangla Sign Language Translation (BSLT) aims to bridge communication gaps for the deaf and mute community. Our approach involves using Mediapipe Holistic to gather key points, LSTM architecture for data training, and Computer Vision for realtime sign language detection with an accuracy of 94%. Keywords=Recurrent Neural Network, LSTM, Computer Vision, Bangla font.

Ismot Sadik Peyas, Zahid Hasan, Md. Rafat Rahman Tushar et al.

In warehouses, specialized agents need to navigate, avoid obstacles and maximize the use of space in the warehouse environment. Due to the unpredictability of these environments, reinforcement learning approaches can be applied to complete these tasks. In this paper, we propose using Deep Reinforcement Learning (DRL) to address the robot navigation and obstacle avoidance problem and traditional Q-learning with minor variations to maximize the use of space for product placement. We first investigate the problem for the single robot case. Next, based on the single robot model, we extend our system to the multi-robot case. We use a strategic variation of Q-tables to perform multi-agent Q-learning. We successfully test the performance of our model in a 2D simulation environment for both the single and multi-robot cases.

R. A. Bindu, A. A. Neloy, S. Alam et al.

This paper presents the design and integration of 3 Survivor, a rough terrain negotiable teleoperated mobile rescue and service robot. 3 Survivor is an improved version of two previously studied surveillance robots named Sigma 3 and Alpha N. In 3 Survivor, a modified double tracked with caterpillar mechanism is incorporated in the body design. A passive adjustment established in the body balance enables the front and rear body to operate in excellent synchronization. Instead of using an actuator, a re configurable dynamic method is constructed with a 6 DOF arm. This dynamic method is configured with the planer, spatial mechanism, rotation matrix, motion control of rotation using inverse kinematics and controlling power consumption of the manipulator using angular momentum. The robot is remotely controlled using a handheld Radio Frequency RF transmitter. 3 Survivor is equipped with a Raspberry Pi 12 MP camera which is used for livestreaming of robot operations. Object detection algorithms are run on the live video stream. The object detection method is built using a Faster RCNN with VGGNet16 architecture of CNN. The entire operations of the robot are monitored through a web control window. Therefore, the control portal provides a brief scenario of the environment to run, control and steer the robot for more precise operation. A very impressive 88.25 percent accuracy is acquired from this module in a rescue operation. Along with the ODM, the sensor system of the robot provides information on the hazardous terrain. The feasibility of the 3 Survivor is tested and presented by different experiments throughout the paper.

R. A. Bindu, A. A. Neloy, S. Alam et al.

This paper introduces a rescue robot named Sigma 3 which is developed for potential applications such as helping hands for humans where a human can not reach to have an assessment of the hazardous environment. Also, these kinds of robot can be controlled remotely with an adequate control system. The proposed methodology forces on two issues : 1. Novel mechanism design for measuring rotation, joints, links of Degree of Freedom DOF for an arm which is integrated with Sigma 3, 2. Precise measuring of end effector motion control over three dimensions. In the proposed mechanism design, the DOF measurement is presented by a planar and spatial mechanism where 4 types of rigid joints build up each DOF with controlling by six High Torque MG996R servo motors. Rotation and DOF measurement are consisting of different theoretical references of Rotation Matrix, Inverse Kinematics with experimental results. Presented methodology over Oscillation Damping performance exhibits less than 3 percent error while configuring for on hands testing. Another evaluation of operating time state strongly defends the mechanism of low power consumption ability.