Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Sim and Real: Better Together
Shirli Di-Castro Shashua, Dotan Di Castro, Shie Mannor
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Train only on simulation samples | DDPG + HER on Fetch Push · Fetch Push, real environment | success rate | Hurt |
| Use mixed sim and real training with q_r=0.1 rather than real-only training | DDPG + HER on Fetch Push · Fetch Push, real environment | real episodes to reach 0.9 success rate: 20K -> ≈10K | Improved |
| Increase the real-data collection probability across intermediate q_r values (ablation) | Mixed sim-real training on Fetch Push · Fetch Push, real environment | success rate | Hurt |
| Set real replay-buffer sampling probability equal to its collection probability (q_r=0.1, beta_r=0.1) (ablation) | Mixed sim-real training on Fetch Push · Fetch Push, real environment | success rate | Hurt |
| Raise beta_r above q_r, to beta_r=0.3, 0.5, or 0.7 (ablation) | Mixed sim-real training on Fetch Push · Fetch Push, real environment | success rate | Improved |
| Set beta_r=0.9 while q_r=0.1 (ablation) | Mixed sim-real training on Fetch Push · Fetch Push, real environment | success rate | Hurt |
| Use Sim-dependent training: start with sim-only, then switch to mixed training after sim success reaches 0.7 (ablation) | DDPG + HER on Fetch Push · Fetch Push, real environment | real episodes to reach 0.9 success rate | Improved |
| Use Sim first: start with sim-only, then switch to real-only after sim success reaches 0.7 (ablation) | DDPG + HER on Fetch Push · Fetch Push, real environment | real episodes to reach 0.9 success rate | Hurt |
| Apply mixed training after transferring a simulation-trained policy to the real environment instead of switching to real-only (ablation) | DDPG + HER on Fetch Push · Fetch Push, real environment | real episodes | Improved |
Closest papers by meaning
- Overcoming the Sim-to-Real Gap: Leveraging Simulation to Learn to Explore for Real-World RL (NeurIPS 2024 · similarity 0.89)
- ASID: Active Exploration for System Identification in Robotic Manipulation (ICLR 2024 · similarity 0.88)
- Position: Scaling Simulation is Neither Necessary Nor Sufficient for In-the-Wild Robot Manipulation (ICML 2024 · similarity 0.87)
- Rapidly Adapting Policies to the Real-World via Simulation-Guided Fine-Tuning (ICLR 2025 · similarity 0.86)
- Learning Interactive Real-World Simulators (ICLR 2024 · similarity 0.86)
- Generalizable Domain Adaptation for Sim-and-Real Policy Co-Training (NeurIPS 2025 · similarity 0.85)
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.
