arxiv_ai 95% Match Research Paper Robotics Researchers,AI Engineers,Control Systems Engineers,Industrial Automation Specialists 1 week ago

Learning to Plan & Schedule with Reinforcement-Learned Bimanual Robot Skills

robotics › manipulation

📄 Abstract

Abstract: Long-horizon contact-rich bimanual manipulation presents a significant challenge, requiring complex coordination involving a mixture of parallel execution and sequential collaboration between arms. In this paper, we introduce a hierarchical framework that frames this challenge as an integrated skill planning & scheduling problem, going beyond purely sequential decision-making to support simultaneous skill invocation. Our approach is built upon a library of single-arm and bimanual primitive skills, each trained using Reinforcement Learning (RL) in GPU-accelerated simulation. We then train a Transformer-based planner on a dataset of skill compositions to act as a high-level scheduler, simultaneously predicting the discrete schedule of skills as well as their continuous parameters. We demonstrate that our method achieves higher success rates on complex, contact-rich tasks than end-to-end RL approaches and produces more efficient, coordinated behaviors than traditional sequential-only planners.

Authors (4)

Weikang Wan

Fabio Ramos

Xuning Yang

Caelan Garrett

Submitted

October 29, 2025

arXiv Category

cs.RO

arXiv PDF

Key Contributions

Introduces a hierarchical framework for long-horizon bimanual manipulation that integrates skill planning and scheduling, enabling simultaneous skill invocation. It uses RL to train primitive skills and a Transformer planner to schedule discrete skills and continuous parameters, achieving higher success rates and more efficient coordination than existing methods.

Business Value

Enables more sophisticated and efficient robotic automation for complex assembly and manipulation tasks in manufacturing and logistics, potentially increasing productivity and reducing errors.

Paper Metadata

Innovation Type

Framework Development

Deployment Feasibility

Moderate, requires significant simulation infrastructure for training and careful integration with physical robots.

Limitations Addressed

The difficulty of long-horizon, contact-rich bimanual manipulation, which requires complex coordination beyond simple sequential actions. Existing end-to-end RL approaches and traditional sequential methods are less effective.

Performance Gains

Achieves higher success rates on complex, contact-rich tasks and produces more efficient, coordinated behaviors than end-to-end RL and traditional sequential approaches.

Technical Tags

bimanual robot skillslong-horizon manipulationcontact-rich taskshierarchical frameworkskill planningschedulingReinforcement Learning (RL)Transformer plannersimulationGPU-accelerated

Research Topics

Robotic ManipulationBimanual CoordinationReinforcement Learning for RoboticsTask Planning and SchedulingSim-to-Real Transfer

Methods & Architectures

Hierarchical FrameworkSkill Planning and SchedulingReinforcement Learning (RL) for primitive skillsTransformer-based PlannerGPU-accelerated Simulation TransformerReinforcement Learning Agents

Applications & Tasks

Robotics Industrial Automation Assembly Tasks Complexity of long-horizon bimanual manipulationNeed for coordination between armsLimitations of purely sequential decision-making Learning to plan and schedule robot skillsExecuting complex bimanual manipulation tasksAchieving efficient and coordinated robot behaviors

Related Fields

RoboticsReinforcement LearningMachine LearningControl SystemsComputer Vision

Keywords

roboticsbimanual manipulationskill learningplanningschedulingreinforcement learningtransformersimulationcontact-richlong-horizonautomation

Academic Context

#Robotic Manipulation#Bimanual Coordination#Reinforcement Learning for Robotics#Task Planning and Scheduling#Sim-to-Real Transfer

Technology Stack

Frameworks & Libraries

Transformer

ML Infrastructure

GPU-accelerated simulation

Commercial Potential

Potential Products

Advanced robotic arms for complex assemblyAutomated manufacturing systemsRobotic assistants for intricate tasks

Target Industries

ManufacturingAutomotiveAerospaceLogisticsElectronics Assembly

Use Case Examples

Robots performing intricate assembly of electronic componentsAutomated systems for handling delicate objects in warehousesRobots assisting in surgical procedures requiring precise manipulation

Competitive Edge

Offers a more effective approach to complex bimanual manipulation by integrating planning and scheduling with RL-learned skills, outperforming existing end-to-end RL and sequential methods.

Market Opportunity

Large and growing market for industrial automation and advanced robotics.

Revenue Models

Licensing of robotic control softwaresale of integrated robotic systems.

Resource Requirements

Compute Needs

High compute requirements for GPU-accelerated simulation and RL training.

Data Requirements

Requires simulation environments and potentially real-world data for skill learning and planning.

Deployment Constraints

Sim-to-real gap, safety considerations for physical robots, complexity of integrating planning and control.

Scalability

The hierarchical approach and simulation-based training aim to improve scalability to new tasks and environments.

Regulatory Considerations

Safety standards for industrial robots

Production Readiness

Maturity Level

Research

Time to Market

3-5 years for widespread industrial adoption.

Patent Potential

High, for the hierarchical framework, planning/scheduling approach, and RL skill learning methods.

View Full Paper Back to Papers