Research map / Reinforcement learning / Imitation learning
Model Predictive Adversarial Imitation Learning for Planning from Observation
Tyler Han, Yanda Bao, Bhaumik Mehta, Gabriel Guo, Sanghun Jung, Anubhav Vishwakarma, Emily Kang, Rosario Scalise, Jason Liren Zhou, Bryan Xu, Byron Boots
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 |
|---|---|---|---|
| deploy the learned reward through an MPPI planner instead of a policy (IRL-MPC vs GAIL) (ablation) | GAIL reward and value · Real-Sim-Real navigation | mean cross-track error (m): 0.56 -> 0.37 | Improved |
| train reward and value end to end with the online planner (MPAIL vs IRL-MPC) (ablation) | IRL-MPC · Real-Sim-Real navigation | mean cross-track error (m): 0.37 -> 0.17 | Improved |
| replace the AIL policy with an online MPPI planner | GAIL · simulated navigation, out-of-distribution initial poses (40x40 m) | final reward | Improved |
| replace the AIL policy with an online MPPI planner | GAIL · simulated navigation task, interaction efficiency | interactions to optimality | Improved |
| learn the dynamics model online (MPAIL OM) instead of using an approximate prior model (ablation) | MPAIL · simulated cartpole task | sample efficiency | Hurt |
| use MPAIL with a learned dynamics model instead of a policy network (ablation) | GAIL and BC · OOD navigation evaluation | final reward | Improved |
| increase MPPI iterations from 2 to 5 at horizon H=10 (ablation) | MPAIL · navigation wall clock time vs GAIL | relative training time: about 10% faster than GAIL -> about 10% slower | Hurt |
| use discount factor eta below 1 on costs in MPPI horizon (ablation) | MPAIL · simulated navigation task | convergence | Hurt |
| add temperature decay to MPPI (ablation) | MPAIL · simulated navigation task | training stability | Improved |
| use AIRL reward instead of GAIL | policy-based AIL · simulated navigation task (multimodal expert data) | convergence | Hurt |
Closest papers by meaning
- Deep Imitative Models for Flexible Inference, Planning, and Control (ICLR 2020 · similarity 0.90)
- Imitation Learning from Observation with Automatic Discount Scheduling (ICLR 2024 · similarity 0.90)
- Rethinking Inverse Reinforcement Learning: from Data Alignment to Task Alignment (NeurIPS 2024 · similarity 0.90)
- Imitation Learning by Reinforcement Learning (ICLR 2022 · similarity 0.90)
- GO-DICE: Goal-Conditioned Option-Aware Offline Imitation Learning via Stationary Distribution Correction Estimation (AAAI 2024 · similarity 0.89)
- Learning from Demonstrations via Capability-Aware Goal Sampling (NeurIPS 2025 · similarity 0.89)
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.
