Suggestions
Ayush Agrawal
Motion Planning and Control for Legged Robots
Ayush Agrawal is a Ph.D. student at UC Berkeley conducting research in the Hybrid Dynamic Robotics Lab under the guidance of Prof. Koushil Sreenath. His work primarily focuses on Nonlinear feedback control for underactuated and hybrid systems, combining skills from fields such as nonlinear and optimal controls, reinforcement learning, and robotics. His expertise involves developing controllers and planning algorithms for various types of robots including bipedal and quadrupedal robots, as well as lower limb exoskeletons. Ayush's educational background includes a Ph.D. in Mechanical Engineering from UC Berkeley, a Master of Science (M.S.) in Mechanical Engineering from Carnegie Mellon University, and a Bachelor of Technology (B.Tech.) in Mechanical Engineering from National Institute of Technology Karnataka. Prior to his Ph.D., Ayush held roles in various prestigious organizations, including Senior Software Engineer at Tesla, Software Engineer II at Agility Robotics, and AI Research Intern at Facebook AI. Additionally, he accumulated experience as a Graduate Student Researcher and Graduate Teaching Assistant at both UC Berkeley and Carnegie Mellon University, as well as roles as an Undergraduate Research Assistant and Intern at Universität Paderborn and Bosch Engineering and Business Solutions, respectively.
Highlights
Stop by our #ICRA2022 presentation tomorrow where we talk about our recent work on quadrupedal locomotion on discrete terrain with visual feedback.
ICRA session: Legged Robots I, 11:00-11:05 on Wednesday 5/25, Room 118A, WeA09.11
Paper: https://lnkd.in/g5HXc546 Video: https://lnkd.in/gJQjZayR
We use trajectory optimization to create an offline motion library. Online, we use visual feedback to find feasible footholds in the scene, and chose a desired trajectory from the motion library, which is tracked by a geometric model predictive controller.
#robotics #leggedrobotics
Excited to share our recent work with Shuxiao Chen, Akshara Rai and Koushil Sreenath on Vision-Aided Dynamic Quadrupedal Locomotion on Discrete Terrain.
Paper: https://lnkd.in/g5HXc546 Video: https://lnkd.in/gJQjZayR
We use trajectory optimization to create an offline motion library. Online, we use visual feedback to find feasible footholds in the scene, and chose a desired trajectory from the motion library, which is tracked by a geometric model predictive controller.
