generated from eliahuhorwitz/Academic-project-page-template
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
577 lines (528 loc) · 27.1 KB
/
index.html
File metadata and controls
577 lines (528 loc) · 27.1 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<!-- Meta tags for social media banners, these should be filled in appropriately as they are your "business card" -->
<!-- Replace the content tag with appropriate information -->
<meta name="description" content="A Lean Ecosystem for Robot Learning at Scale">
<meta property="og:title" content="Robot Control Stack: A Lean Ecosystem for Robot Learning at Scale" />
<meta property="og:description" content="A Lean Ecosystem for Robot Learning at Scale" />
<meta property="og:url" content="https://robotcontrolstack.github.io/" />
<!-- Path to banner image, should be in the path listed below. Optimal dimensions are 1200X630-->
<meta property="og:image" content="static/images/rcs_eye_candy_final.jpg" />
<meta property="og:image:width" content="1500" />
<meta property="og:image:height" content="1500" />
<!-- <meta name="twitter:title" content="TWITTER BANNER TITLE META TAG">
<meta name="twitter:description" content="TWITTER BANNER DESCRIPTION META TAG"> -->
<!-- Path to banner image, should be in the path listed below. Optimal dimensions are 1200X600-->
<!-- <meta name="twitter:image" content="static/images/your_twitter_banner_image.png">
<meta name="twitter:card" content="summary_large_image"> -->
<!-- Keywords for your paper to be indexed by-->
<meta name="keywords" content="RCS">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Robot Control Stack</title>
<link rel="icon" type="image/x-icon" href="static/images/favicon.ico">
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="stylesheet" href="static/css/bulma.min.css">
<link rel="stylesheet" href="static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="static/css/bulma-slider.min.css">
<link rel="stylesheet" href="static/css/fontawesome.all.min.css">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="static/css/index.css">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="https://documentcloud.adobe.com/view-sdk/main.js"></script>
<script defer src="static/js/fontawesome.all.min.js"></script>
<script src="static/js/bulma-carousel.min.js"></script>
<script src="static/js/bulma-slider.min.js"></script>
<script src="static/js/index.js"></script>
</head>
<body>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="title">
<img src="static/images/rcs_logo_line.svg" alt="Project Logo"
style="height: 80px; vertical-align: middle;">
<br><br>A Lean Ecosystem for Robot Learning at Scale
</h1>
<div class="is-size-5 publication-authors">
<!-- Paper authors -->
<span class="author-block">
Tobias Jülg<sup>*1</sup>,</span>
<span class="author-block">
Pierre Krack<sup>*1</sup>,</span>
<span class="author-block">
Seongjin Bien<sup>*1</sup>,</span>
<span class="author-block">
Yannik Blei<sup>1</sup>,</span>
<span class="author-block">
Khaled Gamal<sup>1</sup>,</span>
<span class="author-block">
Ken Nakahara<sup>2</sup>,</span>
<span class="author-block">
Johannes Hechtl<sup>1,3</sup>,</span>
<span class="author-block">
Roberto Calandra<sup>2</sup>,</span>
<span class="author-block">
Wolfram Burgard<sup>1</sup> and </span>
<span class="author-block">
Florian Walter<sup>1,4</sup></span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><br><sup>1</sup>University of Technology Nuremberg,</span>
<span class="author-block"><sup>2</sup>TU Dresden,</span>
<span class="author-block"><sup>3</sup>Siemens AG,</span>
<span class="author-block"><sup>4</sup>Technical University of Munich</span>
<span class="eql-cntrb"><small><br><sup>*</sup>Equal Contribution</small></span>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- ArXiv abstract Link -->
<span class="link-block">
<a href="https://arxiv.org/abs/2509.14932" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>
</span>
<!-- Arxiv PDF link -->
<span class="link-block">
<a href="https://arxiv.org/pdf/2509.14932" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>PDF</span>
</a>
</span>
<!-- Supplementary PDF link -->
<!-- <span class="link-block">
<a href="static/pdfs/supplementary_material.pdf" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Supplementary</span>
</a>
</span> -->
<!-- Github link -->
<span class="link-block">
<a href="https://github.com/RobotControlStack/robot-control-stack" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>
</span>
<!-- Doc link -->
<span class="link-block">
<a href="https://robotcontrolstack.org" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-book"></i>
</span>
<span>Docs</span>
</a>
</span>
<!-- Model Link. -->
<span class="link-block">
<a href="https://huggingface.co/RobotControlStack"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<img src="static/images/hf_icon.svg" />
</span>
<span>Data</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Teaser video-->
<section class="hero teaser">
<div class="container is-max-desktop">
<div class="hero-body">
<video id="tree" autoplay muted loop height="100%" preload="metadata">
<source src="static/videos/grid.mp4" type="video/mp4">
</video>
<h2 class="subtitle has-text-centered">
<!-- RCS: A lightweight layered architecture integrates core tools from robotics and exposes them through a Gymnasium-based API that enables seamless switching between simulation and real-world experiments.
The individual layers can be easily customized in both Python and C++, making RCS equally suitable for end-to-end policy learning and low-level controller design. -->
</h2>
</div>
</div>
<!-- End teaser video -->
<!-- Paper abstract -->
<section class="section hero is-light">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Abstract</h2>
<div class="content has-text-justified">
<p>
Vision-Language-Action models (VLAs) mark a major shift in robot learning.
They replace specialized architectures and task-tailored components of expert policies with large-scale
data collection and setup-specific fine-tuning.
In this machine learning-focused workflow that is centered around models and scalable training,
traditional robotics software frameworks become a bottleneck, while robot simulations offer only limited
support for transitioning from and to real-world experiments.
In this work, we close this gap by introducing <i>Robot Control Stack</i> (RCS), a lean ecosystem
designed from the ground up to support research in robot learning with large-scale generalist policies.
At its core, RCS features a modular and easily extensible layered architecture with a unified interface
for simulated and physical robots, facilitating sim-to-real transfer.
Despite its minimal footprint and dependencies, it offers a complete feature set, enabling both
real-world experiments and large-scale training in simulation.
Our contribution is twofold: First, we introduce the architecture of RCS and explain its design
principles.
Second, we evaluate its usability and performance along the development cycle of VLA and RL policies.
Our experiments also provide an extensive evaluation of Octo, OpenVLA, and Pi Zero on multiple robots
and shed light on the benefits of simulated data for robotic foundation models.
</p>
</div>
</div>
</div>
</div>
</section>
<!-- End paper abstract -->
<!-- 3-MIN VIDEO -->
<section class="section" id="video-3min">
<div class="container is-max-desktop">
<h2 class="title is-3 has-text-centered">RCS in 3 Minutes</h2>
<div class="columns is-centered">
<div class="column is-10">
<div class="video-container"
style="position:relative;padding-bottom:56.25%;height:0;overflow:hidden;border-radius:12px;box-shadow:0 10px 30px rgba(0,0,0,0.1)">
<!-- Replace source with your exported narration video -->
<video controls preload="metadata" style="position:absolute;top:0;left:0;width:100%;height:100%">
<source src="static/videos/rcs_3min_overview.mp4" type="video/mp4">
</video>
</div>
<br>
<p>
Traditional robotics is built around hardware, with many interacting parts and specialized AI modules.
With machine learning taking the lead, this relationship flips around: <em> robots are components of a machine learning pipeline</em>.
</p>
<br>
<p>
Many libraries embrace this and adopt a Python- and ML-first approach, but they often lack robust robotics features and hardware support.
Robust policies require careful debugging in both simulation and hardware, which relies on classical robotics tools.
</p>
<br>
<p>
RCS bridges this gap by combining an ML-first design with the essential robotics tools.
It gives you the means to debug interfaces, validate tasks, and test directly on hardware—while remaining a lightweight pip-installable package with minimal dependencies.
</p>
</div>
</div>
</div>
</section>
<!-- existing frameworks often dont combine the learning part with the robot part or are hard to setup -->
<!-- Arch -->
<section class="section hero is-small">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column">
<div class="hero-body has-text-justified">
<h2 class="title is-3 has-text-centered">Architecture</h2>
<p>
<b>C++/Python API</b>
We provide device APIs in C++ with automatically generated Python bindings, ensuring mirrored functionality in both languages.
A new device can be integrated into RCS in either C++ or in Python, ensuring broad hardware compatibility.
</p>
<br>
<p>
<b>Composable scenes</b>
Higher-level abstractions are built on top of our own device APIs.
They leverage Gymnasium wrappers to enable modular scene creation through composition.
</p>
<br>
<p>
<b>Layered architecture</b>
Because we build upon a minimal low-level device API, you can quickly get up and running with new hardware: implement our interface, benefit from all the wrappers and apps higher up in the stack.
</p>
<br>
<img src="static/images/rcs_architecture_small.svg" alt="RCS Architecture." width="100%">
<div class="content has-text-justified">
Fig. 1: Applications (teleoperation, RL, VLA) interface with the environment (sim or real) through a unified
Gymnasium API. Sensors, actuators, and observers wrap the environment, mutating action/observation
spaces.
</div>
</div>
</div>
</div>
</div>
</section>
<!-- End Arch -->
<!-- TODO: explain architecture -->
<!-- Setups and twins -->
<section class="section">
<div class="container is-max-desktop">
<h2 class="title is-3 has-text-centered">Robot Setups with Digital Twins</h2>
<p>
We evaluate the usability of RCS's hardware oriented features by integrating multiple setups with different robots, grippers, cameras and touch sensors.
In total, four robots, four end-effectors, two cameras and a tactile sensor are implemented, both in simulation and on physical hardware.
</p>
<br>
<style>
/* minimal CSS */
#setup-carousel {
overflow: hidden;
}
/* prevents peeking slides */
#setup-carousel .caption {
/* move caption below dots */
margin-top: 0.5rem;
margin-bottom: 2.75rem;
/* adjust if your dots are taller/shorter */
}
</style>
<div id="setup-carousel" class="carousel">
<div class="item-1 has-text-centered">
<div style="display:grid;grid-template-columns:1fr 1fr;gap:.5rem;align-items:start;justify-items:center;">
<img src="static/images/robots/fr3_setup.jpg" alt="FR3 physical setup"
style="max-width:100%;height:auto;max-height:220px;border-radius:6px;">
<img src="static/images/robots/fr3_twin.png" alt="FR3 digital twin"
style="max-width:100%;height:auto;max-height:220px;border-radius:6px;">
</div>
<p class="is-size-6 caption"><strong>FR3</strong> + Franka Hand; wrist & side cameras.</p>
</div>
<div class="item-2 has-text-centered">
<div style="display:grid;grid-template-columns:1fr 1fr;gap:.5rem;align-items:start;justify-items:center;">
<img src="static/images/robots/xarm7_setup.jpg" alt="xArm7 physical setup"
style="max-width:100%;height:auto;max-height:220px;border-radius:6px;">
<img src="static/images/robots/xarm7_twin.png" alt="xArm7 digital twin"
style="max-width:100%;height:auto;max-height:220px;border-radius:6px;">
</div>
<p class="is-size-6 caption"><strong>xArm7</strong> + Tilburg Hand; side camera.</p>
</div>
<div class="item-3 has-text-centered">
<div style="display:grid;grid-template-columns:1fr 1fr;gap:.5rem;align-items:start;justify-items:center;">
<img src="static/images/robots/ur5e_setup.jpg" alt="UR5e physical setup"
style="max-width:100%;height:auto;max-height:220px;border-radius:6px;">
<img src="static/images/robots/ur5e_twin.png" alt="UR5e digital twin"
style="max-width:100%;height:auto;max-height:220px;border-radius:6px;">
</div>
<p class="is-size-6 caption"><strong>UR5e</strong> + Robotiq 2F-85; wrist & overhead cameras.</p>
</div>
<div class="item-4 has-text-centered">
<div style="display:grid;grid-template-columns:1fr 1fr;gap:.5rem;align-items:start;justify-items:center;">
<img src="static/images/robots/so101_setup.jpg" alt="SO101 physical setup"
style="max-width:100%;height:auto;max-height:220px;border-radius:6px;">
<img src="static/images/robots/so101_twin.png" alt="SO101 digital twin"
style="max-width:100%;height:auto;max-height:220px;border-radius:6px;">
</div>
<p class="is-size-6 caption"><strong>SO101</strong> + built-in gripper; wrist & side cameras.</p>
</div>
</div>
</div>
</section>
<!-- End setups and twins -->
<!-- TODO: Applications -->
<section class="section hero is-small">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-justified">
<div class="column">
<div class="hero-body">
<h2 class="title is-3 has-text-centered">Applications</h2>
<br>
<p>
All implemented robots can be teleoperated with multiple devices and can be used to record data.
We also verify that RCS integrates cleanly into ML pipelines, both in the imitation learning and reinforcement learning settings.
We deploy multiple VLAs, and solve a simple simulated pick-up task with PPO, using proprioceptive and RGB states as observations.
</p>
<br>
<br>
<hr>
<h3 class="title is-4 has-text-centered">Teleoperation & Data Collection</h3>
<!-- First row: 3 videos -->
<div class="columns">
<div class="column">
<h4 class="title is-5">HTC Vive</h4>
<video autoplay muted loop preload="metadata" width="100%">
<source src="static/videos/fr3_teleop.mp4" type="video/mp4">
</video>
</div>
<div class="column">
<h4 class="title is-5">Meta Quest 3</h4>
<video autoplay muted loop preload="metadata" width="100%">
<source src="static/videos/xarm7_teleop.mp4" type="video/mp4">
</video>
</div>
<div class="column">
<h4 class="title is-5">SpaceMouse</h4>
<video autoplay muted loop preload="metadata" width="100%">
<source src="static/videos/ur5e_spacemouse.mp4" type="video/mp4">
</video>
</div>
</div>
<!-- Second row: 2 videos -->
<div class="columns is-centered">
<div class="column is-5">
<h4 class="title is-5">Leader-Follower</h4>
<video autoplay muted loop preload="metadata" width="75%">
<source src="static/videos/so101_teleop.mp4" type="video/mp4">
</video>
</div>
<div class="column is-5">
<h4 class="title is-5">Scripted Data Collection</h4>
<video autoplay muted loop preload="metadata" width="75%">
<source src="static/videos/fr3_scripted.mp4" type="video/mp4">
</video>
</div>
</div>
<br>
<hr>
<h3 class="title is-4 has-text-centered">Reinforcement Learning</h3>
<!-- 2 videos -->
<div class="columns is-centered">
<div class="column is-5">
<video autoplay muted loop preload="metadata" width="75%">
<source src="static/videos/fr3_rl_graph.mp4" type="video/mp4">
</video>
</div>
<div class="column is-5">
<video autoplay muted loop preload="metadata" width="75%">
<source src="static/videos/fr3_rl_robot.mp4" type="video/mp4">
</video>
</div>
</div>
<br>
<hr>
<h3 class="title is-4 has-text-centered">VLA Inference</h3>
<!-- 2 videos -->
<div class="columns is-centered">
<div class="column is-5">
<h4 class="title is-5">Real</h4>
<video autoplay muted loop preload="metadata" width="75%">
<source src="static/videos/fr3_vla2.mp4" type="video/mp4">
</video>
</div>
<div class="column is-5">
<h4 class="title is-5">Simulation</h4>
<video autoplay muted loop preload="metadata" width="75%">
<source src="static/videos/fr3sim_vla.mp4" type="video/mp4">
</video>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- End Applications -->
<!-- Results -->
<section class="section hero is-small">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column">
<div class="hero-body has-text-justified">
<h2 class="title is-3 has-text-centered">Results</h2>
<p>We demonstrate how RCS supports VLA research by investigating VLA generalization across multiple embodiments and assessing the benefit of simulated data for robotic foundation models.
</p>
<br>
<!-- Left column: Image -->
<div class="columns is-vcentered">
<div class="column is-half">
<figure class="image">
<img src="static/images/results/robot_comparison.svg" alt="Descriptive alt text" width="90%">
</figure>
</div>
<!-- Right column: 2x2 grid of videos -->
<div class="column is-half">
<div class="columns is-multiline">
<div class="column is-half">
<video autoplay muted loop playsinline preload="metadata" width="100%">
<source src="static/videos/fr3_vla.mp4" type="video/mp4">
</video>
</div>
<div class="column is-half">
<video autoplay muted loop playsinline preload="metadata" width="100%">
<source src="static/videos/xarm7_vla.mp4" type="video/mp4">
</video>
</div>
<div class="column is-half">
<video autoplay muted loop playsinline preload="metadata" width="100%">
<source src="static/videos/ur5e_vla.mp4" type="video/mp4">
</video>
</div>
<div class="column is-half">
<video autoplay muted loop playsinline preload="metadata" width="100%">
<source src="static/videos/so101_vla.mp4" type="video/mp4">
</video>
</div>
</div>
</div>
</div>
<div class="content has-text-justified">
Fig. 2: We fine-tune Pi Zero on four datasets from different setups.
Each dataset contains fewer than 150 episodes.
The fine-tuned models are deployed on the corresponding setups.
The robots that are more prominent in the base model's data mix achieve better results.
</div>
<img src="static/images/results/sim_real_eval.svg"
alt="Success rate plot over training checkpoints." width="100%">
<div class="content has-text-justified">
Fig. 3: We investigate the impact of simulated data on VLA performance.
Our setup is replicated in simulation and used to generate 500 trajectories using a scripted policy, which is then used to complement our manually collected dataset of 143 trajectories.
The plots show the success rate of the policy, both in the simulated scene and on the hardware, as training progresses.
Success rates in simulation correlate with success rates on the physical robot—consistent with a good evaluation metric.
Adding simulated data to the training mix improves performance in both settings.
</div>
</div>
</div>
</div>
</div>
</section>
<!-- TODO: explain results -->
<!-- End Results -->
<!--BibTex citation -->
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>@inproceedings{juelg2026robotcontrolstack,
title={{Robot Control Stack}: {A} Lean Ecosystem for Robot Learning at Scale},
author={Tobias J{\"u}lg and Pierre Krack and Seongjin Bien and Yannik Blei and Khaled Gamal and Ken Nakahara and Johannes Hechtl and Roberto Calandra and Wolfram Burgard and Florian Walter},
year={2026},
booktitle={Proc.~of the IEEE Int.~Conf.~on Robotics \& Automation (ICRA)},
note={Accepted for publication.}
}
</code></pre>
</div>
</section>
<!--End BibTex citation -->
<footer class="footer">
<div class="container">
<div class="columns is-centered">
<div class="column is-8">
<div class="content has-text-centered">
Website template borrowed from <span class="dnerf"><a
href="https://github.com/nerfies/nerfies.github.io">Nerfies</a></span>.
</div>
</div>
</div>
</div>
</footer>
<!-- Default Statcounter code for Robot Control Stack https://robotcontrolstack.github.io/ -->
<script type="text/javascript">
var sc_project = 13169805;
var sc_invisible = 1;
var sc_security = "64083f10";
</script>
<script type="text/javascript" src="https://www.statcounter.com/counter/counter.js" async></script>
<noscript>
<div class="statcounter"><a title="free hit
counter" href="https://statcounter.com/" target="_blank"><img class="statcounter"
src="https://c.statcounter.com/13169805/0/64083f10/1/" alt="free hit counter"
referrerPolicy="no-referrer-when-downgrade"></a></div>
</noscript>
<!-- End of Statcounter Code -->
</body>
</html>