Doctoral defence: Houman Masnavi ”Visibility Aware Navigation”

On 28 August at 14:15 Houman Masnavi will defend his doctoral thesis ”Visibility Aware Navigation” for obtaining the degree of Doctor of Philosophy ( in Physical and Information Technology)

Supervisiors:

Associate Professor Arun Kumar Singh, University of Tartu
Associate Professor Karl Kruusamäe, University of Tartu

Opponent:

Associate Professor Roel Stephan Pieters, University of Tampere (Finland)

Summary:

Motion planning is an important part of robotics and automation. It involves figuring out the best sequence of movements or actions for a robot to achieve a specific goal while following certain rules. This thesis focuses on a particular type of motion planning called Visibility Aware Navigation (VAN). VAN is used when we want the robot to move between two points or follow a path while always keeping certain objects in its field of view. VAN has various applications in different areas. For example, in target tracking and aerial cinematography, VAN helps ensure that the robot can always see its target while avoiding obstacles. In state estimation and localization, VAN helps improve accuracy by planning paths that maintain visibility of markers like fiducial markers or beacons. VAN can also enable cooperation between larger ground robots and smaller aerial robots. The ground robot can guide the aerial robot as long as they can see each other. However, there are two main challenges with VAN. First, new algorithms need to be developed that can handle complex situations in environments cluttered with obstacles. Second, the solutions should work quickly on small devices that can be carried by small robots like drones. This thesis proposes new motion planning algorithms specifically designed for visibility-aware navigation. These algorithms can handle complex environments with both stationary and moving obstacles. They also perform in real-time on small devices with limited resources. The algorithms are extensively tested in both simulations and real-world environments and compared to existing approaches. The proposed algorithms outperform the existing ones in different aspects such as reducing occlusion time, minimizing control effort, and requiring less computation time.

Defence can be also followed in Zoom: https://ut-ee.zoom.us/j/95305881521?pwd=NlQvVzRRVXJYZ0ZSdk55SERQZDhEZz09  (Meeting ID: 953 0588 1521, Passcode: kaitsmine).