Reliable 7-DOF Manipulator Execution via Arm-Angle Selection and Trajectory Retiming

Abstract

Reliable execution of SRS-type seven-degree-of-freedom redundant manipulators with analytical inverse kinematics requires topological continuity and differential feasibility between planning and control. We present IAS-LBR, a two-layer framework combining Interval-based Arm-Angle Selection (IAS) with Look-ahead Bisection Retiming (LBR). IAS maintains feasible arm-angle interval continuity within a selected posture branch under joint limits, while LBR enforces velocity and acceleration constraints through finite-horizon checks and adaptive time scaling. Experiments on a humanoid platform demonstrate stable sparse tracking and near-singularity traversal. In repeated grasp-transfer evaluation, IAS reduces transition abruptness, while LBR raises the success rate to 100% under the reported protocol and eliminates velocity- and acceleration-limit failures. These results support IAS-LBR as an effective execution-layer bridge from planner references to dynamically feasible joint commands.