Task-driven Perception and Manipulation for Constrained Placement of Unknown Objects

Recent progress in robotic manipulation has dealt with the case of previously\nunknown objects in the context of relatively simple tasks, such as bin-picking.\nExisting methods for more constrained problems, however, such as deliberate\nplacement in a tight region, depend more critically on shape information to\nachieve safe execution. This work deals with pick-and-constrained placement of\nobjects without access to geometric models. The objective is to pick an object\nand place it safely inside a desired goal region without any collisions, while\nminimizing the time and the sensing operations required to complete the task.\nAn algorithmic framework is proposed for this purpose, which performs\nmanipulation planning simultaneously over a conservative and an optimistic\nestimate of the object's volume. The conservative estimate ensures that the\nmanipulation is safe while the optimistic estimate guides the sensor-based\nmanipulation process when no solution can be found for the conservative\nestimate. To maintain these estimates and dynamically update them during\nmanipulation, objects are represented by a simple volumetric representation,\nwhich stores sets of occupied and unseen voxels. The effectiveness of the\nproposed approach is demonstrated by developing a robotic system that picks a\npreviously unseen object from a table-top and places it in a constrained space.\nThe system comprises of a dual-arm manipulator with heterogeneous end-effectors\nand leverages hand-offs as a re-grasping strategy. Real-world experiments show\nthat straightforward pick-sense-and-place alternatives frequently fail to solve\npick-and-constrained placement problems. The proposed pipeline, however,\nachieves more than 95% success rate and faster execution times as evaluated\nover multiple physical experiments.\n

Paper

Similar papers

© 2026 NYSGPT2525 LLC