carpevi/MDP2_Review_2: Dual-Arm Trajectory Planning for Collision-Free Pick-and-Place Operations of Moving Objects
Year: 2026
Type: software
Abstract: This work tackles the challenge of automating the pick-and-place operation of multiple moving objects by a dual-arm robot while avoiding collisions between manipulators and reducing the execution time. Within each formulated two-piece subproblem, the system generates collision-free trajectories that complete the subtask in the minimum number of time steps. This application is addressed from the mathematical framework of Markov Decision Processes (MDP). To do so, the workspace has been discretized to get a finite set of states connected by a set of actions. This work is focused on automation of industrial production and computes the plans by dynamic programming (Value Iteration) on a fully known, deterministic MDP. The result is a methodology for using MDPs in time-changing processes which, when applied to bimanual manipulation, provides two advantages: it accounts for collision avoidance between manipulators and automatically generates robot code compatible with different manufacturers. In planning-level simulations under identical conditions, the proposed planner sorted more pieces than a nearest-piece baseline (73.2% versus 68.0%) and more than an MDP without overlapping subproblems (71.3%). With speed control, all pieces to be sorted were placed in the RobotStudio test cases, and physical experiments confirmed feasibility under the tested conditions.
Author: carpevi
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7223125798