REAP: Learning General Rearrangement Heuristics from Motion-Planner Feedback
AcceptedDaniel Swoboda, Hector Geffner
Conference on Robot Learning (CoRL) · 2026
Robotic rearrangement tasks require planning sequences of actions that fulfill high-level specifications while also being geometrically feasible. Symbolic abstractions make the combinatorial structure of rearrangement tasks tractable, but they usually do not fully capture the geometric constraints that determine which actions a motion policy can realize. This is due to the complexity of expressing the underlying geometric feasibility constraints in the symbolic model. We propose REAP (Rearrangement with Elicited Action Preconditions), a framework in which a relational graph neural network heuristic is learned over a simple symbolic abstraction. A motion policy serves as a supervisor that elicits the hidden geometric preconditions of the problem during training. Deployed with greedy best-first search, the learned heuristic generalizes out-of-distribution along object count, plan length, and object arrangement. Compared to classical heuristic search and TAMP baselines on the same instances, REAP requires approximately an order of magnitude fewer motion policy queries because the heuristic anticipates which low-level actions are geometrically feasible. We show that the same architecture, training recipe, and search procedure adapt to various robotic rearrangement domains.
Full abstract
RearrangementTask and Motion PlanningLearned HeuristicsGeneralised PlanningRobotics
WorkBenchMark: A LEGO-Based Assembly Benchmark with an Assembly-by-Disassembly Baseline for the Smart Manufacturing League
AcceptedWenbo Ma, Daniel Swoboda, Matteo Tschesche, Till Hofmann
RoboCup Symposium · Oral Presentation · 2026
A LEGO Duplo–based benchmark for robotic assembly, motivated by the RoboCup Smart Manufacturing League, where low-level manipulation must meet task-level symbolic reasoning under physical constraints — a combination today's end-to-end methods handle unreliably. It contributes 400 tasks across four difficulty tiers and an open-vocabulary, assembly-by-disassembly baseline whose planning-based pipeline beats a modern vision-language-action approach at every tier. The benchmark, simulator, and baseline will be released openly.
Full abstract
Robotic AssemblyTask and Motion PlanningPlanningBenchmarkAutonomous Robotics
Making Robots Play by the Rules: The ROS 2 CLIPS-Executive
Tarik Viehmann, Daniel Swoboda, Samridhi Kalra, Himanshu Grover, Gerhard Lakemeyer
18th International Conference on Agents and Artificial Intelligence (ICAART) · 2026
Brings CLIPS — a rule-based language well suited to knowledge-driven robot coordination — into the ROS 2 ecosystem as a CLIPS-Executive, building on the original from the Fawkes framework. Its flexibility is shown by layering a PDDL-based planning framework on top.
Full abstract
CLIPSROS 2Rule-Based ReasoningRobot ExecutivePDDL
From Production Logistics to Smart Manufacturing: The Vision for a New RoboCup Industrial League
Supun Dissanayaka, Alexander Ferrein, Till Hofmann, Kosuke Nakajima, Mario Sanz-Lopez, Jesus Savage, Daniel Swoboda, Matteo Tschesche, Wataru Uemura, Tarik Viehmann, Shohei Yasuda
2025 RoboCup Symposium, Salvador, Bahia · 2025
Sets out the vision for the RoboCup Smart Manufacturing League, successor to the Logistics League. It widens the competition from production logistics into a full modern-factory scenario built from initially independent tracks — assembly, human-robot collaboration, humanoid robotics — that gradually merge, keeping the league relevant to current industrial-robotics challenges and welcoming to both newcomers and veteran teams.
Full abstract
RoboCupSmart ManufacturingIndustrial RoboticsCompetition Design
CAD-Based Product Partitioning for Automated Disassembly Sequence Planning with Community Detection
Soren Munker, Daniel Swoboda, Karim El Zaatari, Nehel Malhotra, Lucas Manasses Pinheiro de Souza, Amon Goppert, Chi-Guhn Lee, Robert H Schmitt
Proceedings of the Changeable, Agile, Reconfigurable and Virtual Production Conference · 2023
Addresses disassembly sequence planning for sustainable remanufacturing, such as recovering lithium-ion cells from automotive battery packs. A CAD-based method partitions a product using community detection to derive disassembly sequences with fewer steps, helping make remanufacturing scalable.
Full abstract
Disassembly PlanningCADCommunity DetectionSustainabilityRemanufacturing
