Dexterous manipulation, especially dexterous grasping, is a primitive and critical capability of robots that allows the realization of performing human-like behaviors. However, enabling robots to achieve human-level dexterity still remains a highly challenging problem. This is not only due to the requirement to adapt to the inevitable uncertainties caused by unstructured and dynamic environments, but also due to the many other factors that influence grasping performance, such as the manipulators (parallel grippers, multi-finger hands, etc.), the perception approach and the grasping learning strategy. Traditional robot grasping relies on a set of hypotheses, using simple mathematical models to manage the problem and combining them with some proposed metrics to assess the quality of the grasp. However, such simple, well-defined models cannot be easily applied to complicated scenarios. The purpose of this research project is to report on relevant robot grasping techniques, with a focus on dexterous grasping and manipulation of robots.

In this Research Topic, we will focus on an open, challenging and hot Research Topic: robotic dexterous grasping and manipulation, where a robot is required to understand, as well as adapt to, complex scenarios and grasp different known or unknown objects. Due to the influence of miscellaneous lighting conditions and the intricate spatial positions of objects, most objects, especially whose 3D shapes are relatively sophisticated, will probably have an insurmountable gap between the detected appearance and the real situation. Classical grasping methods include generating the analytic model, which need large amount of prior knowledge or modeling parameters of objects with poor generalization. In recent years, new approaches combining with artificial intelligence, such as learning-based methods, provide new directions for solving this problem. For example, vision-based machine learning (including object localization, object pose estimation and grasp estimation, etc.) is a promising approach to accomplish dexterous robot grasping and manipulation in complex scenarios, with the ability to generalize to handle new objects. Other directions include the design of new mechanisms and end effectors for grasping.

Topics of interest include, but are not limited to, the following:
• Planning and control for robotic dexterous grasping
• Robotic perception in dexterous grasping and manipulation
• Robotic mechanisms of robotic end effectors and hand
• Multi-sensor systems for dexterous grasping and manipulation
• Humanoid dexterous hand design for grasping
• Learning and adaptation for dexterous grasping in cluttered scenes
• Datasets and virtual environments for dexterous grasping and manipulation
• Continuous learning for dexterous grasping and manipulation
• Applications and systems of dexterous grasping and manipulation
• Interdisciplinary research (such as neural mechanisms about dexterous manipulation)

Submit your idea

Impact

Frontiers in Robotics and AI

Participating Journals

Impact