RESEARCH

Designing controllers for Unmanned Aerial vehicles (UAVs) including quadrotors and Autonomous Ground Vehicles (AGVs), are challenging regarding control design, model representation, and stability. Various control approaches have been employed to address the control system limitations of UAVs/AGVs and enhance the stability, guidance, and navigation.

Have a look at our contributions
[Paper4] , [Paper3] , [Paper2] , [Paper1]

In the field of autonomous navigation, comprehensive autonomous modules able to accurately estimate Unmanned Aerial Vehicle (UAV) motion components and to provide control signals to successfully track the vehicle along the desired trajectory are in great demand. When using a Vision-Aided Inertial Navigation System (VA-INS) composed of a low-cost Inertial Measurement Unit (IMU) and a vision unit (monocular or stereo camera), the UAV motion components that require estimation will include orientation (attitude), gyro bias (if angular velocity is unavailable), position, and linear velocity. Given the fact that vehicle’s attitude, position, and linear velocity are generally unknown, they can be reconstructed utilizing sensor measurements. In our Lab, we develop observer-based controllers for Unmanned Aerial Vehicles (UAVs) and Autonomous Ground Vehicles (AGVs).

Visit our research Talks/Videos at international events
Hashim – Observer-based Controller for Attitude – Romania

Have a look at our contributions
[Paper3] , [Paper2] , [Paper1]

Robust and accurate navigation solutions for autonomous vehicles are essential. Indoor and outdoor applications, such as household cleaning devices, pipelines, terrain mapping, reef monitoring, exemplify situations when GPS might be unreliable and only low-cost measurement units (e.g., inertial measurement unit (IMU)) might be available. In such a case GPS-independent navigation solutions are indispensable. A typical low-cost IMU module is composed of an accelerometer and a gyroscope which provide measurements of rigid-body’s acceleration and angular velocity, respectively. In the absence of GPS, a cost-effective autonomous vehicle requires navigation solutions that rely on low-cost IMU and feature measurements collected by a vision unit. Hence, autonomous navigation in space requires estimation of orientation (known as attitude), position, and linear velocity. An inertial vision unit composed of a stereo vision unit and an IMU can be employed to extract rigid-body’s pose – a combination of attitude and position. In our Lab, we develop stochastic and deterministic nonlinear estimation techniques for vehicles navigating with six degrees of freedom (6 DoF) applicable for Unmanned Aerial Vehicles (UAVs) and Autonomous Ground Vehicles (AGVs).

Visit our research Talks/Videos at international events
Akos – Mobile Robots Navigation – Japan
Ajay – Navigation Deterministic Filter – USA
Hashim – Navigation Stochastic Observer – USA

Have a look at our contributions
[Paper4] , [Paper3] , [Paper2] , [Paper1]

Robotics applications are experiencing a surge in demand for navigation solutions suitable for partially or completely unknown robot pose in three-dimensional (3D) space (i.e., attitude and position) within an unknown environment. Simultaneous Localization and Mapping (SLAM) is one of the key robotics tasks as it tackles simultaneous mapping of the unknown environment defined by multiple landmark positions and localization of the unknown pose (i.e., attitude and position) of the robot in 3D space. In our Lab, we develop stochastic and deterministic nonlinear estimation techniques for SLAM applicable for Unmanned Aerial Vehicles (UAVs) and Autonomous Ground Vehicles (AGVs).

Visit our research Talks/Videos at international events
Marium – SLAM Stochastic Filter – USA
Trevor – SLAM Observer – Romania

Have a look at our contributions
[Paper6] , [Paper5] , [Paper4] , [Paper3] , [Paper2] , [Paper1]

Automated and semi-automated robotic applications such as unmanned aerial vehicles (UAVs), autonomous underwater vehicles (AUVs), Autonomous Ground Vehicles (AGVs), satellites, radars and others can be controlled to rotate successfully in the three dimensional (3D) space if the orientation of the rigid-body is accurately known. However, the true attitude (orientation) or pose (orientation + position) of a rigid-body, cannot be extracted directly. Alternatively, the attitude/pose can be determined using a set of measurements available in the body-frame and observations in the inertial-frame. In general, measurement units are corrupted with unknown bias and noise components. In our Lab, we develop stochastic and deterministic nonlinear estimation techniques for attitude/pose applicable for vehicles in 3D space.

Visit our research Talks/Videos at international events
Moise – QUEST-based Kalman filter and LQR – Luxembourg
Hashim – Neural Stochastic Attitude Filter – USA
Akos – Pose Estimation Mobile Robots – Japan
Trenton – Pose Filter – Canada
Ajay – Attitude Estimation – Hong Kong

Have a look at our contributions
[Paper10] , [Paper9] , [Paper8] , [Paper7] , [Paper6] , [Paper5] , [Paper4] , [Paper3] , [Paper2] , [Paper1]

The use of collaborative autonomous robotic vehicles allows for greater flexibility and capacity as well as higher performance in areas such as surveillance, inspection, space explorations, communication, sensor deployment and many others. Multi-agent systems (MAS) distribute work in a logical manner and exchange information via self-formed local network and, hence, they are often called nodes. The network is named a communication graph formed by a set of nodes and the communication lines between different nodes are called edges. The graph can be directed or undirected. An undirected graph allows the information to flow in both directions. The connected nodes of such a graph own similar characteristics. On a directed graph or a digraph the direction of the information flow is fixed. The direction is pointed from one node to another indicating how the information flows from one node to its neighbors. The control of such multi-agent systems faces several practical as well as theoretical challenges. Dynamics of the node can be nonlinear and unknown, the network bandwidth capacity is limited and may suffer from variable delays and loss of packets, the operating environment is changing and complex with presence of noise, the embedded computational resources are limited, etc. In our Lab, we develop multi-agent consensus, formation, and distributed control techniques for homogeneous heterogeneous systems addressing unknown high order nonlinear system.

Have a look at our contributions
[Paper3] , [Paper2] , [Paper1]

Our lives rely heavily on the resources that forests provide. They are regarded as the planet’s lungs because they filter the air by adding oxygen (O2) and lowering the high levels of carbon dioxide levels (CO2). They serve as homes for a variety of animals and can be utilized to shield crops from the wind. Additionally, they clear the water of the majority of pollution-causing agents. Due to the numerous jobs and higher revenues that forests create, countries’ economies are improved. Smoke and air pollution pose serious health threats, especially to vulnerable populations. Evacuations disrupt livelihoods and cause psychological trauma. Economically, the costs are staggering. Firefighting expenditures soar, and losses in timber, agriculture, and tourism industries mount. Long-term, diminished soil fertility hinders agriculture, and reduced water quality impacts communities downstream. Environmental repercussions extend globally. We address critical challenges in wildfire detection, focusing on enhancing time resolution and optimizing processing speeds while maintaining high accuracy levels of state-of-the-art machine learning algorithms.

Have a look at our contributions

[J30]. A. V. Jonnalagadda, and H. A. Hashim, “SegNet: A Segmented Deep Learning based Convolutional Neural Network Approach for Drones Wildfire Detection,” Remote Sensing Applications: Society and Environment (RSA-SE), pp. 1-26, 2024.

We have strong expertise on applying a variety of Artificial Intelligence (AI) techniques to address different engineering applications.

Supervised and/or Unsupervised Learning for Cast Components: [Paper1]

Fuzzy Logic Control for Twin Rotor MIMO System: [Paper1] , [Paper2]

Evolutionary Techniques for Communication Systems: [Paper4] , [Paper3] , [Paper2] , [Paper1]