Robotic Decision Making via Diffusion Models

SVG Icon Editor

Author: Tzu-Yuan Huang

Machine learning for robot decision making

In recent years, an increasing amount of research1 has focused on enabling robots to perform diverse tasks in complex environments using machine learning (ML)-based techniques. Major technology companies, such as NVIDIA and Tesla, are also advancing efforts to introduce service robots capable of extensive human interaction in daily life. The ability to make decisions based on task requirements and environmental changes has become a crucial factor to fully integrate into human life.

ML-based approaches, such as reinforcement learning and imitation learning, leverage training data to implicitly learn task requirements, as well as the dynamics of both the system and its environment, ultimately deriving adaptive strategies for various scenarios. However, these methods encounter significant challenges, particularly in terms of model training stability and their capacity to learn multimodal behaviors.

Diffusion models in robotic decision making

Deep generative models (DGMs) have demonstrated remarkable success in natural language processing and image generation, highlighting their potential for robot policy learning. Among the family of DGMs, diffusion models3 have been widely adopted in robotics, including trajectory planning4, control6, and grasping generation5, owing to their training stability and capability for long-horizon generation.

The core idea behind diffusion models is an iterative denoising procedure, where the neural network learn how to guide the samples from noise distribution to data distribution. In the forward phase, Gaussian noise is gradually injected to a clean data sample so the data is perturbed, while the neural network is trained to denoise and reconstruct the original sample in the reverse process, as depicted in Figure 1. Similar to image generation, the robot trajectory could also be learned by diffusion models and denoised to a task-performed path during generation.

DDPM figure
Fig. 1. Forward process and reverse process in diffusion models. Adapted from: Ho et al. (2020), Denoising Diffusion Probabilistic Models3.

The challenge of diffusion-based decision making

Even though diffusion models have shown potential in robotic decision making due to the strong capability to learn high dimensional behavior, they still face several critical challenges in robotic decision-making. These include, but are not limited to:

  1. Real-time application challenges: The limited inference speed hinders practical deployment.
  2. Lack of safety guarantees: Diffusion-based policies do not inherently ensure safety.
  3. Generalization limitations: These models struggle to handle out-of-distribution scenarios beyond the training data.

Addressing these challenges remains an open research problem2, and its resolution could play a key role in enabling robots to seamlessly integrate into human environments.

Reference

[1] Ravichandar, Harish, et al. "Recent advances in robot learning from demonstration." Annual review of control, robotics, and autonomous systems 3.1 (2020): 297-330. (https://doi.org/10.1146/annurev-control-100819-063206)
[2] Huang, Tzu-Yuan, et al. "SAD-Flower: Flow Matching for Safe, Admissible, and Dynamically Consistent Planning." arXiv preprint arXiv:2511.05355 (2025). (https://doi.org/10.48550/arXiv.2511.05355)
[3] Ho, Jonathan, Ajay Jain, and Pieter Abbeel. "Denoising diffusion probabilistic models." Advances in neural information processing systems 33 (2020): 6840-6851. (https://doi.org/10.48550/arXiv.2006.11239)
[4] Janner, Michael, et al. "Planning with diffusion for flexible behavior synthesis." arXiv preprint arXiv:2205.09991 (2022). (https://doi.org/10.48550/arXiv.2205.09991)
[5] Urain, Julen, et al. "Se (3)-diffusionfields: Learning smooth cost functions for joint grasp and motion optimization through diffusion." 2023 IEEE international conference on robotics and automation (ICRA). IEEE, 2023. (https://doi.org/10.48550/arXiv.2209.03855)
[6] Huang, Tzu-Yuan, et al. "Toward near-globally optimal nonlinear model predictive control via diffusion models." arXiv preprint arXiv:2412.08278 (2024). (https://doi.org/10.48550/arXiv.2412.08278)

About the Author

Tzu-Yuan Huang, M.Sc. is a research associate and doctoral candidate at the Chair of Information-Oriented Control (ITR), Technical University of Munich (TUM), under the supervision of Prof. Sandra Hirche. As a member of relAI, his research focuses on data-driven control and safe, constraint-satisfying generative models for robotic systems, including his recent work on diffusion- and flow-matching-based planning.

RELATED

  • / Valentyn Melnychuk

    Can We Trust the Uncertainty of Causal Foundation Models?

    Causal foundation models promise fast, flexible treatment-effect estimation from observational data, but can their uncertainty be trusted? This post examines prior-data fitted networks (PFNs) for causal inference, revealing a hidden prior-induced confounding bias, and introduces a one-step posterior correction that restores reliable, frequentist-consistent uncertainty for average treatment effect estimation. Full post
  • / Molly Kennedy

    Exploring XAI Methods for Interpretability of Large Language Models

    Large Language Models can summarise text, answer questions, and classify sentiment with impressive accuracy, but their reasoning often feels like a black box. Why did the model land on that prediction? This post walks through four complementary approaches from Explainable AI (XAI) that help open up the box. Together, these methods make LLMs less opaque and more trustworthy. Full post
  • / Lukas Gosch

    A Beginner’s Guide to Certifiable Robustness

    Machine learning models, such as ChatGPT and those used in autonomous driving, are becoming essential tools in our daily lives. However, the existence of "adversarial examples" demonstrates that these systems are not free from vulnerabilities. The post, introduces the concept of adversarial examples and discusses Certifiable Robustness, a methodology designed to combat them. Full post
  • / Balian He

    Responsible Textual Generative Models (Part I): Generating Truthful Content

    This post introduces hallucination in large language models—why fluent outputs can still be wrong, what distinguishes intrinsic from extrinsic errors, and how training data, model design, inference, and alignment contribute. It surveys detection and mitigation approaches and sets up later posts on toxicity, bias, and inclusivity in responsible textual generative AI. Full post
  • / Aswathi

    Random Convolutions: A Simple Way to Boost Generalization

    Random Convolutions (RC) are a simple and effective data augmentation technique designed to improve the robustness of deep learning models, particularly in medical imaging. Instead of relying on fixed or learned filters, RC applies a small stack of randomly generated convolutional filters to input images during training. This introduces strong but controlled variations in image texture while preserving the underlying anatomical structure. By exposing models to a wide range of plausible appearance changes, Random Convolutions reduce overfitting to dataset-specific characteristics such as scanner type or acquisition protocol. In practice, RC acts as an implicit form of domain randomization, encouraging models to focus on stable, shape-based and contextual features rather than fragile texture cues. Despite its simplicity and low computational cost, RC has been shown to substantially improve generalization across institutions and imaging conditions in challenging tasks such as multiple sclerosis lesion segmentation. Full post
  • / Ahmed Abdelrahman

    Neuromorphic Computing: A Brain-inspired Approach to Robot Intelligence

    This post offers a high-level walkthrough of neuromorphic computing, why it matters for energy-efficient robot intelligence, and how SNNs, event cameras, and neuromorphic chips fit together Full post
  • / Shengqiang Zhang

    Introduction to Embodied Instruction Following

    This work explores how AI systems bridge the gap between understanding human instructions and performing real actions by learning through direct interaction with their environment. Full post
  • / Maximilian Fleissner

    From Unlucky Strikers to Statistical Learning Theory

    Using the analogy of football striker, this post explains how statistical learning theory helps us understand when we can trust model predictions and empirical averages, highlighting the challenges of generalization in modern machine learning, especially with complex models like deep neural networks. Full post
  • / Unai Fischer Abaigar

    Performative Prediction

    Machine learning systems are increasingly used to support decision-making processes. Yet, these systems do not merely reflect the world—they also reshape it. Once deployed, predictions can influence behaviors, alter policies, and redirect resources, creating feedback loops that change the very data-generating processes they aim to model. Full post
  • / Moritz Knolle

    What even is differential privacy?

    A concise introduction to differential privacy which offers provably privacy guarantees for training machine learning models. Full post
  • / Sameer Ambekar

    Mitigating Domain shifts

    Adapting a deep neural network to unseen data and tasks is imperative these days, however access to target data is often available. Common target adaptation techniques including domain adaptation and generalization train for meaningful representations during source training. Recent paradigms such as Test-time training/adaptation focus on optimizing the source model on unseen data. To do so, they finetune the model on the streaming unsupervised data which is useful for practical scenarios. Moreover, these techniques can be applied to variety of tasks such as regression, classification and segmentation. Full post
  • / Lisa Wimmer

    A gentle introduction to uncertainty quantification

    Uncertainty Quantification (UQ) is considered indispensable for predictive models in safety-critical applications. Modern models, though high-performing, struggle with providing meaningful uncertainty estimates due to a number of reasons. Full post
  • / Blog Editorial Team

    Welcome to the relAI Blog

    Welcome to the relAI blog of the Konrad Zuse School of Excellence in Reliable AI (relAI). This blog will serve as a platform to share cutting-edge research and developments from our school, highlighting the significant strides we are making towards making AI systems safer, more trustworthy, and privacy-preserving. Full post
LOAD MORE