Research map / Reinforcement learning / Visual and representation RL
A latent control model for realistic rodent motion
Aidan Sirbu, Charles Y Zhang, Yuanjia Yang, Bence Olveczky, Talmo Pereira, Blake Richards
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes an embodied latent control model for generating realistic rodent motion in a virtual animal body, evaluated against imitation-trained baselines (abstract only; details not available).
Problem
Embodied models of animal behavior need to move like real animals to enable fair comparisons between biological and artificial agents; prior virtual rodent work trains realistic motion via imitation.
Approach
A latent control model for realistic rodent motion, building on prior virtual animal bodies and imitation-based motion training; specific mechanism not stated in the available abstract.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add a behavior prior distilled from the frozen encoder, sampled instead of the motion prior (ablation) | SCAMPER · random rollouts, Vendi score | behavioral diversity (Vendi score) | Improved |
| sample behavior prior instead of motion prior during random rollouts (ablation) | SCAMPER · random rollouts, net displacement and arena occupancy | net distance traveled | Improved |
| sample behavior prior instead of motion prior during random rollouts (ablation) | SCAMPER · random rollouts | Kernel Inception Distance to training data | Improved |
| sample behavior prior instead of motion prior (ablation) | SCAMPER · rearing height distribution | Wasserstein distance to training data | Improved |
| distill the behavior prior from the frozen encoder (ablation) | SCAMPER · prior-encoder KL | KL divergence (nats): 14.35 -> 5.88 | Improved |
| use the behavior prior for residual control instead of the motion prior (ablation) | SCAMPER · residual control stability | stability under residual control | Hurt |
| residual control of motion prior instead of MIMIC-MJX | SCAMPER · dense-reward locomotion speed tracking | target speed matching | Improved |
| residual control of motion prior instead of training from scratch | SCAMPER · gait analysis, all 18 metrics | mean distance to real rat gait: walk 3.44 -> 0.59 | Improved |
| residual control of motion prior instead of MIMIC-MJX or scratch | SCAMPER · dense-reward rearing task | task success rate | Improved |
| add behavior prior with learned gating to motion-prior-only variant (ablation) | SCAMPER · sparse-reward arena foraging, large minimum spawn radius | treats collected per episode | Improved |
| add behavior prior on top of motion prior (ablation) | SCAMPER · foraging with small spawn radius | treats collected per episode | No clear effect |
| sampling the motion prior at very high noise (ablation) | motion prior · random rollouts | diversity vs realism | Mixed |
Limitations
- Only the abstract was available; method details and results are unverified.
Experimental setup
Builds on: virtual rodent body (MuJoCo-based)
Closest papers by meaning
- Neural Circuit Architectural Priors for Rat Locomotion (NeurIPS 2026 · similarity 0.89)
- Intrinsic Goals for Autonomous Agents: Model-Based Exploration in Virtual Zebrafish Predicts Ethological Behavior and Whole-Brain Dynamics (NeurIPS 2025 · similarity 0.87)
- Behavioral Exploration: Learning to Explore via In-Context Adaptation (ICML 2025 · similarity 0.86)
- Neural Motion Simulator Pushing the Limit of World Models in Reinforcement Learning (CVPR 2025 · similarity 0.86)
- SEEA-R1: Tree-Structured Reinforcement Fine-Tuning for Self-Evolving Embodied Agents (NeurIPS 2025 · similarity 0.84)
- Self-Improving Embodied Foundation Models (NeurIPS 2025 · similarity 0.84)
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.
