Manipulation Planning for Redundant Robots: A Practical Approach

Abstract

An emerging paradigm in solving the classical motionplanning problem (among static obstacles) is to capture the connectivity of the configuration space using a finite (but possibly large) set of landmarks (or nodes) in it. In this paper, we extend this paradigm to manipulation-planning problem, where the goal is to plan the motion of a robot so that it can move a given object from an initial configuration to a final configuration while avoiding collisions with the static obstacles and other movable objects in the environment. Our specific approach adapts Adriadne's clew algorithm, which has been shown effective for classical motion-planning problems (Mazer et al. 1994;Ahuactzin 1994). In our approach, landmarks are placed in lower dimensional submanifolds of the composite configuration space. These landmarks represent stable grasps that are reachable from the initial configuration. From each new landmark, the planner attempts to reach the goal configuration by executing a local planner, again in a lower (but different) dimensional submanifold of the composite configuration space. The approach is probabilistically resolution complete, does not assume that a closed-form inverse-kinematics solution for the manipulator is available, and is particularly suitable for redundant manipulators. We also demonstrate that our approach is practical for realistic problems in three-dimensional environments with manipulator arms having fairly large numbers of degrees of freedom. We have experimented with this approach for a 7-DOF manipulator in 3-D environments with one movable object, and computation times range between a few minutes and a few tens of minutes-in our experiments, between 3 min to 15 min, depending on the task difficulty.

Planning pre-grasp manipulation for transport tasks

“…A large body of work has focused on modelling the manipulator's configuration space (c-space) [6] in a preprocessing step to find the free-space regions [7] or connected paths for a static environment [8][9][10]. Recent work has also developed several variants of sampling-based planning techniques, such as Rapidlyexploring Random Trees (RRTs) [11], for path-planning in dynamic environments.…”

A General Manipulation Task Planner

“…Let us consider a 3-dimensional workspace with a robot R and a movable object M moving among static obstacles 1 . The robot has n degrees of freedom and M is a rigid object with 6 degrees of freedom that can only move when it is grasped by the robot.…”

“…In [1], the Ariane's Clew algorithm [6] is applied to a redundant robot manipulating a single object in a 3D workspace. The method assumes discrete grasps of the movable object; it is however capable to deal in realistic situations with redundant manipulators for which each grasp possibly corresponds to an infinite number of robot configurations (i.e.…”

“…[1,2,5,12,18]) assume that finite sets of stable placements and possible grasps of the movable object are given in the definition of the problem. Consequently, a part of the task decomposition is thus resolved by the user since the initial knowledge provided with these finite sets has to contain the grasps and the intermediate placements required to solve the problem.…”

MPK: An open extensible motion planning kernel

“…The inclusion of more complex tasks is something that is being actively investigated. Examples include grasping planning, partially specified goal configurations, inverse kinematics, both point-to-point 25 and along a specific tool frame path, and manipulation planning, 26 etc.…”