Does X-VLA run on AgileX PiPER?
Capabilities
| Capability | State | Notes | Evidence |
|---|---|---|---|
| Train / fine-tune | works | Upstream: X-VLA-0.9B fine-tuned on Soft-Fold, 1,200 bimanual AgileX cloth-folding trajectories (released as 2toINF/X-VLA-SoftFold). Community: LeRobot X-VLA checkpoint on single-arm piper_follower data. | ev-xvla-paper-softfold, ev-hub-vis22-xvla |
| Real-robot inference | works with caveats | Upstream: real-world cloth folding on the bimanual AgileX Aloha (about one cloth folded in under two minutes). Through LeRobot on a single Piper: runs but stutters (issue #2942, unresolved). | ev-xvla-paper-softfold-result, ev-xvla-softfold-agilex-readme, ev-issue-2942 |
| Simulation inference | no verified evidence | — | |
| Teleop + record | no verified evidence | — |
Caveats
- Upstream evidence is the bimanual AgileX Aloha platform (two PiPER arms driven through AgileX's Piper ROS 1 Noetic driver) using X-VLA's own server-client deployment, not a single PiPER arm and not LeRobot.
- Through LeRobot, open issue #2942 (two users): X-VLA on a Piper moves intermittently with pauses because action outputs change step-wise; ACT on the same data runs smoothly.
Evidence 5 sources; the cell shows the highest tier
- T1 · upstream documentedpapersource 2025-10-11verified 2026-10-25 (today)
X-VLA paper, Sec. 5: the Soft-Fold cloth-folding dataset is collected on the bimanual AgileX platform and X-VLA-0.9B is fine-tuned on it (Appendix F).
“we build a high-quality cloth-folding dataset on the bi-manual Agilex platform, termed Soft-Fold, which consists of 1,200 trajectories collected through a carefully designed pipeline.”
https://arxiv.org/abs/2510.10274 - T1 · upstream documentedpapersource 2025-10-11verified 2026-10-25 (today)
X-VLA paper abstract: with the 1,200 Soft-Fold demonstrations X-VLA performs real-world dexterous cloth folding; Appendix J lists AgileX among its three real-world embodiments.
“with only 1,200 demonstrations, it excels at a dexterous cloth-folding manipulation task in the real world, achieving an average throughput of folding 1 cloth under two minutes.”
https://arxiv.org/abs/2510.10274 - T1 · upstream documentedofficial-docssource 2025-11-06verified 2026-10-25 (today)
X-VLA repo evaluation/SoftFold-Agilex: serve 2toINF/X-VLA-SoftFold with deploy.py and run the client deploy/client_eef6d_xvla.py against the AgileX Aloha, whose arms are launched with the bundled Piper ROS driver (roslaunch piper start_ms_piper.launch, piper_description URDF).
“We evaluate **X-VLA** on the **Agilex Aloha platform** to perform long-horizon cloth-folding tasks.”
https://github.com/2toinf/X-VLA/blob/6c4ecfb8be458ae7c0c96094e4b0add4174f4fe2/evaluation/SoftFold-Agilex/readme.md - T3 · community reportedhub-checkpointsource 2026-10-11verified 2026-10-24 (1d ago)
config.json type=xvla (base_model lerobot/xvla-base); trained on vis22/green_yellow_plates whose meta/info.json robot_type=piper_follower.
https://huggingface.co/vis22/xvla-piper-robot - T3 · community reportedgithub-issuesource 2026-02-10verified 2026-10-24 (1d ago)lerobot 0.4.3
Bug report: X-VLA controlling a Piper arm produces stuttering motion; a second user reports the same; open, no fix.
“When using the XVLA model to control the Piper robotic arm, the arm's movement exhibits noticeable stuttering (i.e., moving intermittently with pauses).”
https://github.com/huggingface/lerobot/issues/2942
X-VLA inference speed by hardware robot-independent · inference cost only
| Compute | Runtime | Precision | Cams | p50 / p95 ms | Inference | Platform | Source |
|---|---|---|---|---|---|---|---|
| Jetson Orin Nano 8GB (Super) | TensorRT | fp16 | 3 @640x480 | 342.59 / 343.47 | 2.92 Hz | jetpack 7.2.1 · power_mode MAXN_SUPER, clocks pinned with jetson_clocks (GPU 1020 MHz, CPU 1728 MHz) · tensorrt 10.16.2.10conditionsboard: Jetson Orin Nano Engineering Reference Developer Kit Super (8 GB) · l4t: R39.2.1 · os: Ubuntu 24.04.4 LTS · python: 3.12.3 · desktop: running Inference cost only; not task success. 300 s run on synthetic 640x480 observations resized by the exported policy to 224x224; 10 denoising steps; latency is runtime-only over a pre-materialized observation ring. Whole-board RAM in use 3.04 GB (unified memory, desktop running). Actions were gated against the stock policy (cosine >= 0.99999) before timing. | T3 · community reported |
| Jetson Orin Nano 8GB (Super) | ONNX (split) | fp16 | 3 | 391.55 / 407.33 | 2.55 Hz | jetpack 7.2 (L4T R39.2.x, per the repo host-setup doc at this commit) · power_mode MAXN_SUPER, pinned clocks · cuda 13.2 · tensorrt 10.16.2 (system, via ONNX Runtime TensorRT EP)conditionsboard: Jetson Orin Nano Super 8 GB developer kit · os: Ubuntu 24.04 · onnxruntime-gpu: 1.24 Inference cost only; not task success. Deterministic synthetic observations, pinned MAXN_SUPER clocks. Configured EP priority 12 graphs TensorRT. System RAM 5395 MB in use. | T3 · community reported |
| Jetson Orin Nano 8GB (Super) | PyTorch eager | fp32 | 3 | 2313.5 / 2320.89 | 0.43 Hz | jetpack 7.2 (L4T R39.2.x, per the repo host-setup doc at this commit) · power_mode MAXN_SUPER, pinned clocks · cuda 13.2 · lerobot 0.6.1conditionsboard: Jetson Orin Nano Super 8 GB developer kit · os: Ubuntu 24.04 Inference cost only; not task success. Deterministic synthetic observations, pinned MAXN_SUPER clocks. Stock LeRobot X-VLA policy in PyTorch FP32 (LeRobot 0.6.1 env). At 30 Hz control the 30-step chunk is overrun (author's control-rate fit). System RAM 5450 MB in use. | T3 · community reported |
Embed this cell
Badge for READMEs and model cards. It always shows the current state.
[](https://caniembody.pages.dev/compat/agilex-piper/xvla)<iframe src="https://caniembody.pages.dev/embed/agilex-piper/xvla" width="380" height="132" style="border:0" title="X-VLA on AgileX PiPER | CanIEmbody"></iframe>History
- 2026-10-25 fecccd1 Initial release: robot policy compatibility matrix · Danilo Znamerovszkij