Inverse kinematics can for computing a goal configuration for the robot arm. Then, efficient path planning techniques can be used to search the configuration space (C-space) of the arm directly for a collision-free path connecting the inital and the goal configurations.

Due to the high-dimensionality of the space, brute-force search algorithms are infeasible. As an alternative, we have conducted a series of experiments utilizing randomized path planning techniques. In particular, we have developed the RRT-Connect planner based on Rapidly-Exploring Random Trees (RRTs) for quickly solving single-query path planning problems in high-dimensional spaces.

A user can click on a object and command the robot to reposition it interactively. All of the motions for the robot to grasp, move, and release the object are computed "on-the-fly", with no preprocessing of the environment. For most tasks, collision-free motions for a 7-DOF arm can be generated in a few seconds.

This problem formulation is suitable for manipulation tasks involving relatively lightweight objects that can be grasped by a single arm. We are currently extending the technique to handle multi-arm manipulation tasks.

Given a robot's internal model of the environment and a statically-stable desired posture, we use a randomized path planner to search the configuration space of the full body of the robot for a collision-free path that also satisifies the balance constraints.

Instead of sampling the entire configuration space, the subspace corresponding to statically-stable postures of a given support configuration (single-leg or dual-leg). Balance constraints are imposed on incremental search motions in order to maintain the overall dynamic stability of the computed trajectories.

Experiments