# Crowd-FM: Learned Optimal Selection of Conditional Flow Matching-generated Trajectories for Crowd Navigation

arXiv: https://arxiv.org/abs/2602.06698 · HTML: https://mlsanity.com/paper/2602.06698/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models.
- Builds on 15 of its 21 references that are in the map.
- First posted on arXiv 6 Feb 2026.

## Placed in

- Diffusion models / [Diffusion and flow policies for robotics](https://mlsanity.com/diffusion/diffusion-and-flow-policies-for-robotics/)

## Builds on (most cited first, 15 of 15)

- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [Improving and generalizing flow-based generative models with minibatch optimal transport](https://mlsanity.com/paper/2302.00482/) (2023)
- [NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration](https://mlsanity.com/paper/2310.07896/) (2023)
- [Crowd-Robot Interaction: Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning](https://mlsanity.com/paper/1809.08835/) (2018)
- [Learning Robotic Manipulation Policies from Point Clouds with Conditional Flow Matching](https://mlsanity.com/paper/2409.07343/) (2024)
- [Relational Graph Learning for Crowd Navigation](https://mlsanity.com/paper/1909.13165/) (2019)
- [DRL-VO: Learning to Navigate Through Crowded Dynamic Scenes Using Velocity Obstacles](https://mlsanity.com/paper/2301.06512/) (2023)
- [DenseCAvoid: Real-time Navigation in Dense Crowds using Anticipatory Behaviors](https://mlsanity.com/paper/2002.03038/) (2020)
- [FlowNav: Combining Flow Matching and Depth Priors for Efficient Navigation](https://mlsanity.com/paper/2411.09524/) (2024)
- [Learning Sampling Distribution and Safety Filter for Autonomous Driving with VQ-VAE and Differentiable Optimization](https://mlsanity.com/paper/2403.19461/) (2024)
- [GPD: Guided Polynomial Diffusion for Motion Planning](https://mlsanity.com/paper/2501.18229/) (2025)
- [Watch out! There may be a Human. Addressing Invisible Humans in Social Navigation](https://mlsanity.com/paper/2211.12216/) (2022)
- [PRIEST: Projection Guided Sampling-Based Optimization For Autonomous Navigation](https://mlsanity.com/paper/2309.08235/) (2023)
- [CrowdSurfer: Sampling Optimization Augmented with Vector-Quantized Variational AutoEncoder for Dense Crowd Navigation](https://mlsanity.com/paper/2409.16011/) (2024)
- [LV-DOT: LiDAR-visual dynamic obstacle detection and tracking for autonomous robot navigation](https://mlsanity.com/paper/2502.20607/) (2025)

## Built on by (most cited first, 2 of 2)

- [FLASH: Efficient Visuomotor Policy via Sparse Sampling](https://mlsanity.com/paper/2605.15492/) (2026)
- [ECTraj: Enhanced Consistency Training for Multi-Agent Trajectory Prediction](https://mlsanity.com/paper/2605.08572/) (2026)

## Abstract (arXiv)

Safe and computationally efficient local planning for mobile robots in dense, unstructured human crowds remains a fundamental challenge. Moreover, ensuring that robot trajectories are similar to how a human moves will increase the acceptance of the robot in human environments. In this paper, we present Crowd-FM, a learning-based approach to address both safety and human-likeness challenges. Our approach has two novel components. First, we train a Conditional Flow-Matching (CFM) policy over a dataset of optimally controlled trajectories to learn a set of collision-free primitives that a robot can choose at any given scenario. The chosen optimal control solver can generate multi-modal collision-free trajectories, allowing the CFM policy to learn a diverse set of maneuvers. Secondly, we learn a score function over a dataset of human demonstration trajectories that provides a human-likeness score for the flow primitives. At inference time, computing the optimal trajectory requires selecting the one with the highest score. Our approach improves the state-of-the-art by showing that our CFM policy alone can produce collision-free navigation with a higher success rate than existing learning-based baselines. Furthermore, when augmented with inference-time refinement, our approach can outperform even expensive optimisation-based planning approaches. Finally, we validate that our scoring network can select trajectories closer to the expert data than a manually designed cost function.