Kourosh Darvishreceived his B.Sc. and M.Sc. degrees in Aerospace Engineering from K.N. Toosi University of Technology and Sharif University of Technology (Tehran, Iran), in2012 and 2014, respectively. During his bachelor's and master's studies, he worked on dynamics, control, estimation, and system identification techniques. Later, Kourosh completed his PhD in Bioengineering & Robotics from theUniversity of Genoa, Italy in 2019. During his PhD, he developed a hierarchical architecture for flexible human-robot collaboration for factory environments and assembly tasks. Since November 2018, he had been working as a post-doc at the Italian Institute of Technology (IIT) in the Artificial and Mechanical Intelligence (previously named Dynamic Interaction Control) Lab and collaborating on the H2020 European project AnDy. During this project till December 2020, he worked on human perception and humanoid robot teleoperation. Since January 2020, he had been collaborating on the ANA Avatar XPRIZE competition, which aims to create an avatar system that can transport human presence to a remote location in real-time. Since January 1st, 2021, he had started to work on the H2020 European Project SoftManBot, where he led the activities related to task learning and safety enhancement for human-robot collaboration. Moreover, from 2021 to 2022, he worked on the ergoCub project, aimed at enhancing the ergonomics of workers in the workplace. In this project, he developed AI algorithms for online human action recognition and motion prediction for the purpose of robot control in collaborative scenarios.
Kourosh serves as a postdoctoral researcher as a part of the Acceleration Consortium in Robot Visionand Learning (RVL) and People, AI, & Robots (PAIR) labs at the University of Toronto. He works diligently on the integration of different AI techniques with the corpus of robots in order to enable them to naturally interact with humans and make a positive impact on humans' lives.
The research aims to introduce a new generation of human-robot collaboration (HRC) systems for complex manipulation tasks, happening every day in chemistry laboratories. For chemistry experiments, to fill the gap between currently manual manipulation and fully autonomous manipulation, robots should learn extensively from the human domain expertise. However, principal elements of human intelligence to perform a wide variety of manipulation tasks in daily life activities are yet unclear. The research hopes to transfer human intelligence for manipulation of lab equipments and materials to robots employing different AI techniques. It will enable robots toward automatically performing chemistry lab experiments. A chemistry lab workspace is intrinsically uncertain, semi-structured, might be occluded, and partially observable. One can decompose a chemistry lab experiment into two levels, at the task level (i.e., high level) where the experiment unfolds into a sequence of actions to be carried out by humans and robots, and at the motion level where skillfully an atomic manipulation action, as a part of a planned sequence of actions, should be performed by humans or robots. In order to successfully deploy robotic technologies to chemistry labs and allow for a natural, fluent, and safe human-robot collaboration, this research intends to address both task level and motion level challenges at the same time, by empowering robots to learn tasks and acquire necessary skills from human experts. We will combine different AI techniques with rich and various perceptual information to enable the next generation of robots for this purpose.