Grasping Utensils for Cooking

Another bottom-level yet important component within the robotic cooking pipeline is grasping. In order to use a tool during robotic cooking, the robot chef first needs to apply a firm grasping, which is influenced by different grasp planning algorithms, over the tool. We have published papers regarding grasp planning approaches in two main aspects: extraction from demonstration and maximizing task coverage.

Robot grasp planning based on demonstrated grasp strategies
Human has developed the master skill in grasping during years of practicing and learning. In this paper, we show that grasp type and thumb placement, which commonly exist in almost all grasps, provide partial constraints of hand postures and wrist positions as well as orientations. They can be integrated into robotic grasp planning procedure to generate feasible grasp based on target object geometry and manipulation task. We have tested this grasp planning approach on both simulators and real system and the results that our approach tends to generate grasps that are closer to human-style and more robust dealing with perception uncertainties. Our approach can also be generalized to other robotic models by considering dimensionality differences between hand models.

Grasp planning to maximize task coverage
During robotic cooking, the same tool may be used for different purposes. In other words, the grasp planning approach should also consider the task. In this paper we show a task-oriented grasp planning approach computing from a task disturbance based task-oriented grasp quality metric. A manipulation task is modeled by building non-parametric statistical distribution of disturbance from demonstration data. This paper also proposes a task-oriented grasp quality criterion based on the distribution of the task disturbance by computing the ratio of disturbance a grasp covers. We have validated the grasp planning approach in both simulation and real system. The results verify the consistency of the success rate with the proposed disturbance-based quality metric

Publication List

[2] Lin, Yun, and Yu Sun. “Grasp Planning to Maximize Task Coverage.” The International Journal of Robotics Research, vol. 34, no. 9, Aug. 2015, pp. 1195–1210
This paper presents a novel robot grasping planning approach that extracts grasp strategies (grasp type, and thumb placement and direction) from human demonstration and integrates them into the grasp planning procedure to generate a feasible grasp concerning the target object geometry and manipulation task.