--- language: - en license: apache-2.0 tags: - robotics - diffusion-policy - embodied-ai - imitation-learning - visuomotor-control - lerobot - ddim - inverse-kinematics metrics: - accuracy model-index: - name: visuomotor-diffusion-policy-engine results: - task: type: robotics name: Multimodal Robotic Manipulation dataset: name: Robotic Manipulation Trajectory Benchmarks type: At1uri/robotic-manipulation-trajectory-benchmarks metrics: - name: Obstacle Avoidance Success Rate type: accuracy value: 0.980 - name: Push-T Success Rate type: accuracy value: 0.945 --- # Visuomotor Diffusion Policy Engine [![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://opensource.org/licenses/Apache-2.0) [![Hub Type](https://img.shields.io/badge/Hub-Model-green.svg)](https://huggingface.co/At1uri/visuomotor-diffusion-policy-engine) [![Robotic Benchmarks](https://img.shields.io/badge/Dataset-Manipulation_Benchmarks-orange.svg)](https://huggingface.co/datasets/At1uri/robotic-manipulation-trajectory-benchmarks) [![Interactive Studio](https://img.shields.io/badge/Space-Robotics_Studio-red.svg)](https://huggingface.co/spaces/At1uri/robotics-diffusion-policy-studio) ## 📌 Executive Summary **`At1uri/visuomotor-diffusion-policy-engine`** is a Conditional 1D Temporal U-Net **Visuomotor Diffusion Policy** and **Damped Least Squares Inverse Kinematics** engine engineered for high-dimensional, contact-rich robotic manipulation and multimodal obstacle avoidance. Standard deterministic behavioral cloning policies trained with Mean Squared Error (MSE) collapse to the conditional expectation $\mathbb{E}[\mathbf{a} \mid \mathbf{o}_t]$, which lies directly inside invalid collision regions when expert demonstrations span multiple valid topological paths. By generating receding-horizon action chunks $\mathbf{a}_{t:t+H}$ via reverse **Denoising Diffusion Implicit Models (DDIM)** with a cosine noise schedule, this engine accurately samples distinct collision-free modes in **10 deterministic denoising steps** ($11.4\text{ ms}$). --- ## 🔬 Mathematical Formulation ### 1. Conditional Action Denoising (DDIM Receding Horizon) Given observation history $\mathbf{o}_t$, the policy initializes a pure Gaussian action sequence $\mathbf{a}^K \sim \mathcal{N}(\mathbf{0}, \mathbf{I})$ of horizon $H=16$ and iteratively denoises along the cosine schedule $\bar{\alpha}_k$: $$\hat{\mathbf{a}}_0 = \frac{\mathbf{a}^k - \sqrt{1 - \bar{\alpha}_k}\,\boldsymbol{\epsilon}_\theta(\mathbf{a}^k, \mathbf{o}_t, k)}{\sqrt{\bar{\alpha}_k}}$$ $$\mathbf{a}^{k-1} = \sqrt{\bar{\alpha}_{k-1}}\,\hat{\mathbf{a}}_0 + \sqrt{1 - \bar{\alpha}_{k-1}}\,\boldsymbol{\epsilon}_\theta(\mathbf{a}^k, \mathbf{o}_t, k)$$ The score network $\boldsymbol{\epsilon}_\theta$ learns the gradient of the multimodal log-density $\nabla_{\mathbf{a}} \log p(\mathbf{a}_{t:t+H} \mid \mathbf{o}_t)$, ensuring trajectories bifurcate cleanly into valid upper/lower obstacle clearance basins without mode-averaging collapse. ### 2. Damped Least Squares Inverse Kinematics (Levenberg-Marquardt) To map denoised end-effector waypoints $\mathbf{x}_t \in \mathbb{R}^2$ to articulated 3-DOF robotic joint angles $\mathbf{q} \in \mathbb{R}^3$, the kinematic solver minimizes regularized tracking error via the analytical Jacobian $\mathbf{J}(\mathbf{q}) = \frac{\partial f(\mathbf{q})}{\partial \mathbf{q}}$: $$\Delta \mathbf{q} = \mathbf{J}^T \left( \mathbf{J}\mathbf{J}^T + \lambda^2 \mathbf{I} \right)^{-1} (\mathbf{x}_{\text{target}} - f(\mathbf{q}))$$ --- ## ⚡ Benchmark Performance ($N=200$ Test Episodes) | Task Benchmark | Diffusion Policy (DDIM-10) | ACT | MSE Behavioral Cloning | |---|---|---|---| | **Obstacle Avoidance Reach** | **98.0%** | 91.5% | 22.5% (Collision Collapse) | | **Push-T Planar Manipulation** | **94.5%** | 88.0% | 41.0% | | **Pick & Place Cube** | **96.5%** | 94.0% | 68.0% | | **Peg-in-Hole Insertion** | **91.0%** | 85.5% | 44.0% | --- ## 🚀 Quickstart & Inference ```python from diffusion_policy_engine import VisuomotorDiffusionPolicyEngine policy = VisuomotorDiffusionPolicyEngine() # Sample a collision-free receding-horizon action chunk in 10 DDIM steps res = policy.sample_action_chunk_ddim( start_xy=[0.20, -0.35], goal_xy=[0.20, 0.35], obs_xy=[0.20, 0.00], obs_radius=0.12, num_inference_steps=10, seed=42 ) print(f"Selected Mode: {res['selected_mode']} | Clearance: {res['min_obstacle_clearance_m']} m") # Solve 3-DOF Inverse Kinematics for endpoint ik = policy.solve_inverse_kinematics(res["trajectory_xy"][-1]) print(f"Joint Angles (rad): {ik['joint_angles_rad']} | IK Error: {ik['position_error_m']} m") ``` --- ## 🔗 Integrated Ecosystem - **Interactive Robotics Studio**: [At1uri/robotics-diffusion-policy-studio](https://huggingface.co/spaces/At1uri/robotics-diffusion-policy-studio) - **Benchmark Dataset**: [At1uri/robotic-manipulation-trajectory-benchmarks](https://huggingface.co/datasets/At1uri/robotic-manipulation-trajectory-benchmarks) - **Research Code Vault**: [At1uri/robotics-diffusion-policy-vault](https://huggingface.co/buckets/At1uri/robotics-diffusion-policy-vault) - **Flagship Collection**: [Institutional Scientific AI Suite](https://huggingface.co/collections/At1uri/institutional-quantitative-ai-suite-6aeeaed0365774ae33b9621e) --- ## 📜 Citation ```bibtex @misc{atluri2026robopolicy, author = {Atluri, Ramprasad}, title = {Visuomotor Diffusion Policy Engine}, year = {2026}, publisher = {Hugging Face}, howpublished = {\url{https://huggingface.co/At1uri/visuomotor-diffusion-policy-engine}} } ```