Research map / Reinforcement learning / Visual and representation RL
Temporal-Logic-Based Reward Shaping for Continuing Reinforcement Learning Tasks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add temporal-logic-based reward shaping | Differential Q-learning · Continuing tasks overall | average-reward learning rate | Helped |
| Add temporal-logic shaping from kitchen advice | Differential Q-learning · Continual area sweeping, perfect kitchen advice | average reward learning rate | Helped |
| Strictly enforce kitchen advice with shielding | Differential Q-learning · Continual area sweeping, imperfect kitchen advice with corridor rewards | average reward | Hurt |
| Add temporal-logic shaping from kitchen advice | Differential Q-learning · Continual area sweeping, imperfect kitchen advice with corridor rewards | average reward | Helped |
| Use shielding to enforce human-visible advice | Differential Q-learning · Continual area sweeping, human is the only source of trash | average reward | Helped |
| Combine human-visible and corridor potentials | Temporal-logic shaping · Continual area sweeping, human-visible and corridor advice | average reward | Helped |
| Use shielding with advice that excludes the corridor | Differential Q-learning · Continual area sweeping, human-visible and corridor advice | average reward | Hurt |
| Use temporal-logic shaping from the scoring-range specification | Deep differential Q-learning · Continuing cart pole | average reward learning performance | Helped |
| Use shielding based on the inaccurate scoring range | Deep differential Q-learning · Continuing cart pole, inaccurate scoring range | average reward | Hurt |
| Use temporal-logic shaping from down-and-right advice | Differential Q-learning · Continuing grid world, perfect advice | average reward learning performance | Mixed |
| Replace temporal-logic shaping with hand-crafted distance shaping | Differential Q-learning · Continuing grid world | average reward learning performance | No clear effect |
| Use temporal-logic shaping despite advice made inaccurate by a wall | Differential Q-learning · Continuing grid world, imperfect advice | average reward | Helped |
Closest papers by meaning
- Reward Shaping for Reinforcement Learning with An Assistant Reward Agent (ICML 2024 · similarity 0.90)
- Automatic Reward Shaping from Confounded Offline Data (ICML 2025 · similarity 0.90)
- Centralized Reward Agent for Knowledge Sharing and Transfer in Multi-Task Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Reinforcement Learning with LTL and $\omega$-Regular Objectives via Optimality-Preserving Translation to Average Rewards (NeurIPS 2024 · similarity 0.88)
- ORSO: Accelerating Reward Design via Online Reward Selection and Policy Optimization (ICLR 2025 · similarity 0.88)
- Highly Efficient Self-Adaptive Reward Shaping for Reinforcement Learning (ICLR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.