Research map / Reinforcement learning / Visual and representation RL

ToS: Tree-of-Skill Reinforcement Learning driven by Behavior Tree for Long-Horizon Manipulation

NeurIPS 2026 · Reinforcement learning › Visual and representation RL · approach: Goal-conditioned policy learning

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes ToS, a tree-of-skill reinforcement learning framework driven by behavior trees for long-horizon robotic manipulation.

Problem

Long-horizon manipulation requires reliable low-level skills, but skill dependencies (e.g., precise object poses) are hard to model symbolically, and existing skill chaining approaches struggle with them.

Approach

Uses behavior trees both as a modular control architecture and as the driver of reinforcement learning for low-level skills, organizing skills into a tree structure so dependencies between skills are handled during learning rather than only through symbolic planning. Full method details are unavailable beyond the abstract.

Closest papers by meaning

• SkillTree: Explainable Skill-Based Deep Reinforcement Learning for Long-Horizon Control Tasks (AAAI 2025 · similarity 0.90)
• SCaR: Refining Skill Chaining for Long-Horizon Robotic Manipulation via Dual Regularization (NeurIPS 2024 · similarity 0.88)
• Multi-Stage Manipulation with Demonstration-Augmented Reward, Policy, and World Model Learning (ICML 2025 · similarity 0.86)
• Goal-Oriented Skill Abstraction for Offline Multi-Task Reinforcement Learning (ICML 2025 · similarity 0.86)
• Chain-of-Action: Trajectory Autoregressive Modeling for Robotic Manipulation (NeurIPS 2025 · similarity 0.86)
• Robust Policy Learning via Offline Skill Diffusion (AAAI 2024 · similarity 0.86)