Jump to Navigation

Motion Planning and Control for Large Agile Robot Teams

دکتر سمانه حسینی

Samaneh Hosseini is a postdoctoral fellow of Aerospace Engineering at University of Toronto in joint with University of Ryerson and Arrowonics Company. She obtained her B.S. and M.Sc. degrees in computer engineering from Kashan University, (2005) and University of Isfahan (2007) respectively. She received her first Ph.D. degree in Software Engineering from University of Isfahan (Iran, 2012) and the second one in Communications and Information Systems from University of Waterloo (Canada, 2015). Her research interests lie at the intersection of computer science, robotics, and optimization. In particular she studies application of biologically inspired algorithms, evolutionary and optimization methods for control and coordination of large multi-agent systems. The application areas of his research include driverless cars, unmanned aerial vehicles, distributed aerial surveillance systems, sensor networks and many others.

Motion Planning and Control for Large Agile Robot Teams
مسئول برگزاری: محمد حسین منشئی
محل: سالن سمینار 3 دانشکده برق و کامپیوتر / تاریخ: ۱۳۹۷-۰۷-۲۳ / زمان: ۱۲:۳۰ ب.ظ

Recent advances in algorithmic robotics allowed unmanned aerial vehicles to navigate through complex environments using on-board sensing and computational resources. However, only few existing solutions claim that they can simultaneously satisfy all the requirements of real applications such as robustness, scalability, optimality, low computation and communication overhead, being fast, real-time, flexible, etc. In this talk, I present a novel algorithm that attempts to balance these goals. This algorithm enables autonomous robots to generate their individual trajectories independently only knowing the relevant position of neighboring robots. To do so, each robot applies a control input function inherited from the Flocking algorithm and contains two main terms: collision avoidance and navigational feedback. The first term keeps two agents away from each other when they become close and guarantees a separation distance between them. It is also able to keep the agents away from static and dynamic obstacles. The second term attracts each agent toward its goal location. We will show that proposed algorithm is able to find collision free paths with lower transition time without any need to know the velocity of neighbors and with less computational overhead in comparison to ORCA, one of the most famous and efficient algorithms in this field. The performance of the proposed algorithm is examined over several dense and complex 2D and 3D benchmark simulation scenarios. The results show that proposed algorithm computes collision-free paths for all of the scenarios without experiencing any deadlock while outperforming ORCA with respect to both transition and execution time. The algorithm scales well to very large robot teams involving thousands of robots. Our approach can compute safe and smooth trajectories for thousands of agents in dense environments with static and dynamic obstacles in few milliseconds.

Seminars | by Dr. Radut

تحت نظارت وف بومی