Session: Rising Stars of Mechanical Engineering Celebration & Showcase

Paper Number: 148583

148583 - Morphological Computation for Resilient Dynamic Locomotion of Compliant Legged Robots With Application to Precision Agriculture 

The overall goal of this research is to investigate how compliance embedded into a legged robot can be harnessed to facilitate control and computation, with an eye to enabling efficient and resilient navigation in real agricultural fields. Research activities innovate along three key foundational robotics research directions.  We first consider hardware design and dynamic modeling. We seek to offer fundamental insights and develop models regarding the effect of various forms of compliance on center of mass motion and gait stabilization for certain classes of legged robots (primarily quadrupeds). We have introduced a new quadruped with three degrees of freedom per leg and a linear one degree of freedom compliant spine, as well as a transormable wheel-leg robot that can negotiate different terrains and objects often found in agricultural fields. The quadruped features a prismatic spring-loaded spine along with a locking-unlocking mechanism that operates based on developed models that seek to optimize energetic efficiency and aggressive performance (such as high speed and large pitch angle jumping forward).  Second, we focus on locomotion control for such robots. We have established an adaptive compliance-agnostic controller for a single purely soft leg (among other robotic systems). Our approach employs Koopman operator theory for real-time learning of a higher-dimensional model of the original (nonlinear) system, which is then used to determine how a desired robot trajectory should be adjusted to accommodate for effects of uncertainty. We have also developed a central pattern generator-based control to enable our transformable wheel-leg robot to navigate efficiently over a range of engineered and natural unstructured terrains. Third, we develop non-holonomic motion planning and autonomous navigation algorithms in agricultural fields specifically. The main idea is to eventually utilize distinctive features of robot body morphology and embedded compliance for efficiency and resilience during autonomous legged locomotion over real agricultural fields. At this point we have perfomed fully autonomous navigation with standard wheeled robots despite large distrurbances that may occur when operating in the field. We have focused on multimodal sensor integration both for navigation (e.g., by fusing LiDAR, IMU, and stereo camera data) as well as scene understanding (e.g., associating plant health from near-infrared and thermal imaging to their geometric characteristics estimated via LiDAR data). To further stimulate foundational robotics research in precision agriculture, we have made publicly available a large diverse field dataset, and outlined methodologies to create high-fidelity digital twins. Overall, this research can transform the science and technology of autonomous legged robots by making them more efficient and resilient in their operation, and thus unlock legged robots' full potential in precision agriculture.

Presenting Author: Konstantinos Karydis University of California, Riverside

Presenting Author Biography: Konstantinos Karydis is an Associate Professor of Electrical and Computer Engineering with cooperating faculty appointments in the Departments of Mechanical Engineering and Computer Science and Engineering. He is also a core faculty member at the Center for Environmental Research and Technology and the Center for Robotics and Intelligent Systems, a steering committee member for BCOE Robotics Programs as well as the Undergraduate Advisor of the Robotics Engineering BS Program. Karydis received with Honors his undergraduate diploma (combined BS and MS program) in Mechanical Engineering from the National Technical University of Athens, Greece in 2010 (3rd in class of 120+ students). He received his PhD in Mechanical Engineering from the University of Delaware in 2015 (Allan P. Colburn Dissertation Award finalist). Prior to joining UCR in 2017, Karydis was a Postdoctoral Researcher on Robotics in the Department of Mechanical Engineering and Applied Mechanics at the University of Pennsylvania. At UCR, Karydis directs the Autonomous Robots and Control Systems Lab. Karydis’ research program addresses foundational and contemporary robotics problems within an integrated theoretical and computational implementation framework paired with significant experimental testing and validation system deployment, thereby developing complete robotic systems that find applications in precision agriculture, environmental monitoring, emergency response, and human-robot collaboration. His research has attracted significant extramural funding from NSF (including a 2021 NSF CAREER Award) USDA and DoD, as well as UC-internal funds (including a UCR OASIS large project and a UCOP MRPI).

Authors:

Konstantinos Karydis University of California, Riverside