XInfer improves robot fleet serving efficiency under real constraints

Scaling VLA Serving Under Robotic Execution Constraints

RoboticsDistributed, Parallel, and Cluster Computing

Summary

Many robots use vision, language, and action models to understand and manipulate their environment, but running these models efficiently across many robots is challenging. The authors present XInfer, a system that manages these models smartly to save computing resources and ensure robots get quick responses when they need them. XInfer balances running many computations at once without slowing down time-critical robot tasks. Tests show it can handle more robots effectively than current methods. The system is publicly available for others to use and build on.

What this means in practice

  • For robot fleet operators: Increase the number of robots served simultaneously with fewer computing resources by managing model computations efficiently under real-time execution constraints.
  • For industrial automation teams: Improve deployment efficiency of robots in manufacturing or logistics by providing faster robot feedback while running complex models on shared hardware.

Authors

Yue Zhang, Zicong Ye, Lufang Chen, Dance Yang, Mingyang Li, Zhenwei Duan, Michael Zhao, Ziye Deng, Starrick Liu, Ruizhi Chengze, YongHua Chen, SongAo Han, Yu Sun, Ran Yu, Xiaofan Li, Ruyi Gan, Hao Wang

Abstract

Vision-language-action (VLA) models have emerged as a promising paradigm for general-purpose robotic manipulation. In our production setting, hundreds of VLA-powered robots are already operating in real-world environments, creating an immediate need for efficient fleet serving. We present XInfer, a production-deployed VLA serving system designed to reduce the inference infrastructure cost of large-scale robotic deployments. XInfer jointly coordinates heterogeneous VLA computation with the evolving response requirements of robotic execution. By exploiting computational parallelism while managing interference, XInfer improves GPU efficiency while protecting time-critical robotic feedback. We evaluate XInfer across three VLA models using production-derived workloads, achieving 1.26-3.75 times the effective robot capacity (ERC) of state-of-the-art systems under physical-execution constraints. Our implementation is publicly available at https://github.com/X-Square-Robot/xinfer.