NUST Institutions Library Catalogue Search for 'an:"125661"'

Augmented Reality Based SLAM /

By Naveed, Izma. . 77p. ; , Motion planning is crucial for helping autonomous robots navigate complex environments efficiently. Recently, Augmented Reality (AR) has been introduced to improve Human-Robot Interaction (HRI) in mobile robot motion planning. However, AR gap-based reactive control systems often suffer from issues like sensor noise and inaccuracies, leading to higher levels of jerk and stress. On the other hand, Simultaneous Localization and Mapping (SLAM) provides a global understanding of the environment, ensuring robust navigation even in dynamic or unfamiliar areas. In this paper, we propose an AR-based Hector Simultaneous Localization and Mapping (SLAM) method for intuitive indoor mobile robot navigation that reduces jerk and stress. Our approach uses AR to set navigation goals and provide visual markers for the user, while SLAM ensures accurate real-time mapping for precise navigation and obstacle avoidance. For path planning, the robot uses Dijkstra's algorithm for global planning and Trajectory Rollout for local planning. We tested the effectiveness of our AR-based Hector SLAM in three different scenarios and compared the results with an admissible gap-based navigation algorithm. Experimental results showed that our method improved jerk and stress by 11.63% and 11.39% respectively, leading to smoother and safer trajectories. 30cm..

Place Hold on Augmented Reality Based SLAM /

Development of Ethical Framework for Metaverse /

By Ahmed ,Soban. Soft Copy, .

Place Hold on Development of Ethical Framework for Metaverse /

Fast Marching Trees (FMT*) For Dynamic Motion Planning

By Salman Sadiq ,Muhammad . 30cm..

Place Hold on Fast Marching Trees (FMT*) For Dynamic Motion Planning

Gait Generation for a Quadrupedal Robot / Zainullah Khan

By Khan, Zainullah . . 89p. , Quadrupedal robots have gained significant research interest due to their ability to achieve agile and stable locomotion over complex terrains. Such locomotion can be achieved by combining various gaits, however, simply changing robot gaits does not guarantee robust and stable behavior. To ensure stable robot locomotion, gaits must be seamlessly blended. Current methods of gait transition include model-based, mainly Model Predictive Control (MPC), approaches, which are limited by the use of handengineered gaits; Reinforcement Learning (RL)-based methods, which address these limitations but require extensive training; and hybrid methods that combine multiple controllers but still experience abrupt gait timing changes. This thesis introduces a novel RL-MPC hybrid control framework that addresses the controllers’ shortcomings in the current literature. The proposed controller incorporates a feature extractor module that extracts features from the robot terrain and state. The novel framework also introduces a gait timing correction step to smooth out gait transitions. The proposed framework was tested on a randomly generated rough terrain, where the robot efficiently traversed and transitioned between gaits while maintaining accurate command velocity. Testing the effectiveness of the contact timing correction step revealed that the locomotion produced by the controller without contact timing correction was jerky and unstable on rough terrain. The proposed framework also outperforms a state-of-the-art method in gait transitioning, resulting in smoother and more stable locomotion. 30cm.

Place Hold on Gait Generation for a Quadrupedal Robot /

Impact Dynamics for Humanoid Robot

By Jameel, Saad . . 111p. , Research on biped robots focuses on replicating human behavior such as walking, jumping, and kicking. The kicking motion, in particular, poses significant challenges due to the need for precise balance and coordination of joint movements and the optimization of joint variables critical for effective kicking. Existing kicking techniques generally rely on kinematic models and predictive model assumptions without incorporating the full dynamics of the robot. Most models use keyframe-based and Inverse Kinematics (IK)-based techniques for joint trajectories and apply feedback control methods such as Dynamic Movement Primitives (DMP), Zero Moment Point (ZMP) control, and reinforcement learning-based control for stability and walking motion. These methods can produce a kicking motion but do not account for the kicking dynamics. Moreover, these techniques are limited to fully actuated robots. This thesis introduces a dynamically inspired, underactuated biped robot operating in a sagittal plane capable of walking and kicking. The model’s dynamics are derived using the Euler-Lagrange method and controlled through a Hybrid Zero Dynamics (HZD)-based Input-Output Linearization (IOL) strategy to achieve precise trajectory tracking. These trajectories are parameterized by the underactuated joint and optimized via Sequential Quadratic Programming (SQP), ensuring that torque remains within permissible limits. This approach incorporates impact dynamics to maintain stability during the walking and kicking phases. The model’s effectiveness is validated using the NAO robot platform in a 3D physics simulator. Our results demonstrate that the robot executes kicks faster, with an average kicking time of 0.75 seconds, and achieves long-range kicks, with an average kicking distance of approximately 6.1 meters. These capabilities surpass the performance of the current state-of-the-art Q-learning-based kicking engines. 30cm.

Place Hold on Impact Dynamics for Humanoid Robot

Control of Flywheel Inverted Pendulum Using Reinforcement Learning /

By Ahmad, Shakeel . . 67p. , Balancing an inverted pendulum is a classic control problem that traditionally requires precise system modeling for effective controller design. Reinforcement Learning (RL) offers a model-free alternative but requires extensive training, which is impractical and risky when performed directly on physical hardware. Existing methods typically rely on simulation environments built on accurate models, which are often difficult to obtain. In this work, we use RL to balance flywheel inverted pendulum by constructing an approximate model of the system through parameter estimation. Despite its inaccuracies, the model proved sufficient for training RL agents in simulation. We developed a simulation environment based on the estimated model and trained agents using Deep Q-Network (DQN), Proximal Policy Optimization (PPO), and Discrete Soft Actor-Critic (SAC) algorithms. The trained policies were deployed on real hardware without any additional fine-tuning. All agents achieved successful swing-up and stabilization, with SAC achieving the fastest swing-up time (1.65 s) and lowest steady-state error (0.0220 rad), demonstrating that RL can tolerate model imperfections and still perform effectively on real systems. 30cm.

Place Hold on Control of Flywheel Inverted Pendulum Using Reinforcement Learning /

Addressing the Problem of Sloshing in a Liquid Carrying Mobile Robot through Artificial Intelligence /

By Khan, Khubaib Haider . . 90p. , Liquid sloshing presents a critical challenge for mobile robots tasked with transporting partially filled containers, as it destabilizes motion, reduces accuracy, and increases the risk of spillage. Traditional passive and active control methods—such as baffles, PID, and LQR—are limited in adaptability and computational efficiency when faced with nonlinear, time-varying fluid–structure interactions. This thesis addresses these challenges by formulating slosh suppression as a reinforcement learning problem, leveraging the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm integrated with a surrogate ball-in-container analogy in the Webots simulator. The surrogate transforms complex liquid dynamics into a tractable cart-pole inspired system, enabling efficient training of robust control policies without reliance on high-fidelity but computationally expensive CFD models. A custom simulation–learning pipeline was developed, coupling Webots with a PyTorchbased TD3 agent via JSON-based communication. State and action spaces were defined using the ball’s displacement ratio and robot velocity, while a multi-component reward function balanced slosh minimization, forward progress, and smooth energy-efficient motion. Training results demonstrated that the TD3 agent learned to achieve stable, sloshfree navigation over a 10 m trajectory, outperforming traditional controllers and earlier RL variants in stability and adaptability. Results shows that the Model Learns well after 600 Episodes attaining optimal velocity of 1.25 – 0.5 m/s keeping ball within 2 mm limit, maximizing the reward to around 1180 at 1700 Episodes, eventually reducing sloshing by 65 % during Training phase. While, during Test phase robot is tested for 0 to 4 m/s velocity with control and no control scenarios, showing the results RL algorithm suppress the ball displacement (liquid slosh) by approximately 45 %. The study validates reinforcement learning as a viable paradigm for real-time liquid slosh suppression in robotics, offering superior robustness and scalability. Contributions include an AI-driven control framework for slosh suppression, integrationxvii of surrogate modeling with DRL for real-time feasibility, and a reward design framework encoding domain-specific stability objectives. For Further Future working it is recommended to include extending to real liquid models and hardware validation, adopting symmetric action spaces and advanced RL methods, and integrating slosh control with autonomous navigation in dynamic environments. 30cm.

Place Hold on Addressing the Problem of Sloshing in a Liquid Carrying Mobile Robot through Artificial Intelligence /

Drug Discovery Using Generative AI /

By Razzaq, Ayesha . . 104p. , This study investigated application of the Conditional Variational Autoencoder (CVAE) for de novo molecular designing focusing on three targets : Cyclin-dependent kinase 2 (CDK2),Peroxisome Proliferator Activator Receptor-gamma (PPAR-gamma),Dipeptide peptidase 4 (DPP-4). SMILES-based molecular representations is coupled with the physicochemical properties such as molecular weight, logP, hydrogen bond donors/acceptors, TPSA, and rotatable bonds. CVAE is trained to encode the meaningful lower-dimensional latent space representation of compounds. The resulting molecules are also checked for drug-likeness(QED ,SA), novelty,uniqueness and other metrics i.e. binding affinity using computational screening pipelines. SMILES format of the structural outputs were converted to SDF and PDB files and docked against targets in PyRx and binding interactions are analyzed in Discovery Studio. 30cm.

Place Hold on Drug Discovery Using Generative AI /

Physics Inspired Machine Learning Simulator for Heating and Insulation Effects on Houses /

By Saleem, Muhammad Awais . . 102p. , Over 40% of all primary energy use comes from buildings, which also has the potential to provide the grid with some degree of energy flexibility. HVAC systems are crucial for ensuring a comfortable indoor environment for building occupants. While classical white-box models must adhere to the laws of physics, their accuracy may be limited by their inherent structure. Additionally, they have significant computational overhead, making them less practical for testing control systems. Several studies have utilized black-box models to develop simulators that forecast the indoor environment, but these studies typically focus on short-term prediction horizons of up to 1 week. Furthermore, few studies have accounted for the uncertainties in weather conditions, which can significantly impact the indoor environment and lead to energy wastage in HVAC systems. In this paper, we propose the development of a fast and accurate simulator for a multi-zone dwelling that can forecast the indoor environment for 6 months while incorporating weather uncertainties. We used 9 machine learning models namely, LR, Lasso, Ridge, Elastic Net, Extra trees, Regression Trees, Random Forest, KNN, and XGBoost xviii and 1 deep learning model namely, MLP. Hyperparameters are optimized for the machine learning models, and the indoor environment is forecasted. Among the models implemented, Lasso regression emerged as the best model with an RMSE of 0.077 for a 3-month forecast while KNN has an RMSE of 0.09 for 6-months prediction. 30cm.

Place Hold on Physics Inspired Machine Learning Simulator for Heating and Insulation Effects on Houses /