---
license: mit
pipeline_tag: reinforcement-learning
---
# Kinesis target goal reaching (& directional control)
        
## Model Description
- This model was trained on the KIT-Locomotion dataset and finetuned on a downstream target goal reaching task.
- The model performs muscle control without reference trajectories using a musculoskeletal model (MyoLeg).
- For more details, please refer to the [paper](arxiv.org/abs/2503.14637) and the [code repository](https://github.com/amathislab/Kinesis).

## Model Details
- **Framework**: PyTorch
- **Model Type**: MLP
- **Training Data**: KIT-Locomotion dataset & Task-driven RL
- **Input**: Proprioceptive state and target goal
- **Output**: Target muscle lengths