Fast, Anytime Motion Planning for Prehensile Manipulation in Clutter

Abstract

Many methods have been developed for planning the motion of robotic arms for picking and placing, ranging from local optimization to global search techniques, which are effective for sparsely placed objects. Dense clutter, however, still adversely affects the success rate, computation times, and quality of solutions in many real-world setups. The current work integrates tools from existing methodologies and proposes a framework that achieves high success ratio in clutter with anytime performance. The idea is to first explore the lower dimensional end effector's task space efficiently by ignoring the arm, and build a discrete approximation of a navigation function, which guides the end effector towards the set of available grasps or object placements. This is performed online, without prior knowledge of the scene. Then, an informed sampling-based planner for the entire arm uses Jacobianbased steering to reach promising end effector poses given the task space guidance. While informed, the method is also comprehensive and allows the exploration of alternative paths over time if the task space guidance does not lead to a solution. This paper evaluates the proposed method against alternatives in picking or placing tasks among varying amounts of clutter for a variety of robotic manipulators with different end-effectors. The results suggest that the method reliably provides higher quality solution paths quicker, with a higher success rate relative to alternatives.

Sample Complexity of Probabilistic Roadmaps via $ε$-nets

“…While many practical robotic systems are high dimensional, i.e., d 6, some application instances naturally admit decoupling of the degrees of freedom of the system, which induces lower-dimensional configuration subspaces. For instance, manipulation problems (see, e.g., [2]) can be typically decomposed into a sequence of tasks where the manipulator is driving toward an object (while fixing its arms), then moves an arm towards the object, and finally grasps it by actuating its fingers. Additionally, in some settings prior knowledge about the structure of the environment or a lower-dimensional space can inform sampling in the full configuration space, which can lead to more informative sampling distributions [28], [29], [30], [31].…”

“…PRM is particularly suitable in multiquery settings, where the workspace environment needs to be preprocessed to answer multiple queries consisting of different start and goal points. Recently PRM has been applied to challenging robotic settings, including manipulation planning [2], inspection planning and coverage [3], task planning [4], and multi-robot motion planning [5], [6]. PRM is instrumental in many modern single-query planners, which implicitly maintain a PRM graph and return a solution that minimizes the path's cost [7], [8], [9], [10], [11].…”

Comprehensive Review on Reaching and Grasping of Objects in Robotics

“…250 Thus, the real challenge in robots is when a robot gripper needs to plan and navigate extremely cluttered environments. 251 For instance, in contrast to storing, kitting needs to prepare products or tools quickly, and it needs to pick up an object from a cluttered place and to put down an object in cluttered environments via real-time planning, which can exceed the complexity of objects and detect the issues of collision. 252…”

Review of Deep Reinforcement Learning-Based Object Grasping: Techniques, Open Challenges, and Recommendations

“…Evidently, several robotic solutions for the grasping task are associated with human hands [303]. Thus, the real challenge in robots is when the robot gripper needs to plan and navigate in extremely cluttered environments [304]. In contrast to merely storing, kitting needs to prepare products or tools quickly, pick up objects from a cluttered place and place these objects down in cluttered environments by using real-time planning, which can overcome object complexity and detect collision issues [305].…”