Robust Grasping

Robust grasping in robotics aims to enable robots to reliably manipulate objects despite uncertainties in object shape, pose, and environmental conditions. Current research focuses on developing advanced control algorithms, including Bayesian optimization and reinforcement learning, to improve grasp planning and execution, often incorporating tactile and visual feedback for enhanced precision and adaptability. This work leverages diverse approaches such as modular neural networks, bio-inspired designs (e.g., compliant actuators and grippers), and novel 3D object representations (e.g., neural fields) to achieve robust grasping across various object types and manipulation tasks. The resulting advancements have significant implications for applications ranging from industrial automation and assistive robotics to search and rescue operations.