Skip to content

Commit bba4334

Browse files
juelgjuelg
committed
added system performance
1 parent dcbb857 commit bba4334

2 files changed

Lines changed: 1564 additions & 74 deletions

File tree

‎index.html‎

Lines changed: 18 additions & 74 deletions
Original file line numberDiff line numberDiff line change
@@ -3,24 +3,15 @@
33

44
<head>
55
<meta charset="utf-8">
6-
<!-- Meta tags for social media banners, these should be filled in appropriately as they are your "business card" -->
7-
<!-- Replace the content tag with appropriate information -->
8-
<meta name="description" content="A Lean Ecosystem for Robot Learning at Scale">
6+
<meta name="description" content="A lean, ROS-free sim-to-real framework for training and deploying Vision-Language-Action (VLA) models and RL agents. Native MuJoCo Gymnasium wrappers with synchronous execution for Franka, UR5e, xArm, and SO101.">
97
<meta property="og:title" content="Robot Control Stack: A Lean Ecosystem for Robot Learning at Scale" />
108
<meta property="og:description" content="A Lean Ecosystem for Robot Learning at Scale" />
119
<meta property="og:url" content="https://robotcontrolstack.github.io/" />
12-
<!-- Path to banner image, should be in the path listed below. Optimal dimensions are 1200X630-->
1310
<meta property="og:image" content="static/images/rcs_eye_candy_final.jpg" />
1411
<meta property="og:image:width" content="1500" />
1512
<meta property="og:image:height" content="1500" />
1613

1714

18-
<!-- <meta name="twitter:title" content="TWITTER BANNER TITLE META TAG">
19-
<meta name="twitter:description" content="TWITTER BANNER DESCRIPTION META TAG"> -->
20-
<!-- Path to banner image, should be in the path listed below. Optimal dimensions are 1200X600-->
21-
<!-- <meta name="twitter:image" content="static/images/your_twitter_banner_image.png">
22-
<meta name="twitter:card" content="summary_large_image"> -->
23-
<!-- Keywords for your paper to be indexed by-->
2415
<meta name="keywords" content="RCS">
2516
<meta name="viewport" content="width=device-width, initial-scale=1">
2617

@@ -58,7 +49,6 @@ <h1 class="title">
5849
<br><br>A Lean Ecosystem for Robot Learning at Scale
5950
</h1>
6051
<div class="is-size-5 publication-authors">
61-
<!-- Paper authors -->
6252
<span class="author-block">
6353
Tobias Jülg<sup>*1</sup>,</span>
6454
<span class="author-block">
@@ -91,7 +81,6 @@ <h1 class="title">
9181

9282
<div class="column has-text-centered">
9383
<div class="publication-links">
94-
<!-- ArXiv abstract Link -->
9584
<span class="link-block">
9685
<a href="https://arxiv.org/abs/2509.14932" target="_blank"
9786
class="external-link button is-normal is-rounded is-dark">
@@ -101,7 +90,6 @@ <h1 class="title">
10190
<span>arXiv</span>
10291
</a>
10392
</span>
104-
<!-- Arxiv PDF link -->
10593
<span class="link-block">
10694
<a href="https://arxiv.org/pdf/2509.14932" target="_blank"
10795
class="external-link button is-normal is-rounded is-dark">
@@ -113,18 +101,6 @@ <h1 class="title">
113101
</span>
114102

115103

116-
<!-- Supplementary PDF link -->
117-
<!-- <span class="link-block">
118-
<a href="static/pdfs/supplementary_material.pdf" target="_blank"
119-
class="external-link button is-normal is-rounded is-dark">
120-
<span class="icon">
121-
<i class="fas fa-file-pdf"></i>
122-
</span>
123-
<span>Supplementary</span>
124-
</a>
125-
</span> -->
126-
127-
<!-- Github link -->
128104
<span class="link-block">
129105
<a href="https://github.com/RobotControlStack/robot-control-stack" target="_blank"
130106
class="external-link button is-normal is-rounded is-dark">
@@ -134,7 +110,6 @@ <h1 class="title">
134110
<span>Code</span>
135111
</a>
136112
</span>
137-
<!-- Doc link -->
138113
<span class="link-block">
139114
<a href="https://robotcontrolstack.org" target="_blank"
140115
class="external-link button is-normal is-rounded is-dark">
@@ -144,7 +119,6 @@ <h1 class="title">
144119
<span>Docs</span>
145120
</a>
146121
</span>
147-
<!-- Model Link. -->
148122
<span class="link-block">
149123
<a href="https://huggingface.co/RobotControlStack"
150124
class="external-link button is-normal is-rounded is-dark">
@@ -163,22 +137,16 @@ <h1 class="title">
163137
</section>
164138

165139

166-
<!-- Teaser video-->
167140
<section class="hero teaser">
168141
<div class="container is-max-desktop">
169142
<div class="hero-body">
170143
<video id="tree" autoplay muted loop height="100%" preload="metadata">
171144
<source src="static/videos/grid.mp4" type="video/mp4">
172145
</video>
173146
<h2 class="subtitle has-text-centered">
174-
<!-- 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.
175-
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. -->
176-
</h2>
147+
</h2>
177148
</div>
178149
</div>
179-
<!-- End teaser video -->
180-
181-
<!-- Paper abstract -->
182150
<section class="section hero is-light">
183151
<div class="container is-max-desktop">
184152
<div class="columns is-centered has-text-centered">
@@ -209,17 +177,13 @@ <h2 class="title is-3">Abstract</h2>
209177
</div>
210178
</div>
211179
</section>
212-
<!-- End paper abstract -->
213-
214-
<!-- 3-MIN VIDEO -->
215180
<section class="section" id="video-3min">
216181
<div class="container is-max-desktop">
217182
<h2 class="title is-3 has-text-centered">RCS in 3 Minutes</h2>
218183
<div class="columns is-centered">
219184
<div class="column is-10">
220185
<div class="video-container"
221186
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)">
222-
<!-- Replace source with your exported narration video -->
223187
<video controls preload="metadata" style="position:absolute;top:0;left:0;width:100%;height:100%">
224188
<source src="static/videos/rcs_3min_overview.mp4" type="video/mp4">
225189
</video>
@@ -246,9 +210,6 @@ <h2 class="title is-3 has-text-centered">RCS in 3 Minutes</h2>
246210
</div>
247211
</section>
248212

249-
<!-- existing frameworks often dont combine the learning part with the robot part or are hard to setup -->
250-
251-
<!-- Arch -->
252213
<section class="section hero is-small">
253214
<div class="container is-max-desktop">
254215
<div class="columns is-centered has-text-centered">
@@ -283,13 +244,6 @@ <h2 class="title is-3 has-text-centered">Architecture</h2>
283244
</div>
284245
</div>
285246
</section>
286-
<!-- End Arch -->
287-
288-
<!-- TODO: explain architecture -->
289-
290-
291-
292-
<!-- Setups and twins -->
293247
<section class="section">
294248
<div class="container is-max-desktop">
295249
<h2 class="title is-3 has-text-centered">Robot Setups with Digital Twins</h2>
@@ -356,9 +310,6 @@ <h2 class="title is-3 has-text-centered">Robot Setups with Digital Twins</h2>
356310
</div>
357311
</div>
358312
</section>
359-
<!-- End setups and twins -->
360-
361-
<!-- TODO: Applications -->
362313
<section class="section hero is-small">
363314
<div class="container is-max-desktop">
364315
<div class="columns is-centered has-text-justified">
@@ -377,7 +328,6 @@ <h2 class="title is-3 has-text-centered">Applications</h2>
377328
<hr>
378329
<h3 class="title is-4 has-text-centered">Teleoperation & Data Collection</h3>
379330

380-
<!-- First row: 3 videos -->
381331
<div class="columns">
382332
<div class="column">
383333
<h4 class="title is-5">HTC Vive</h4>
@@ -399,7 +349,6 @@ <h4 class="title is-5">SpaceMouse</h4>
399349
</div>
400350
</div>
401351

402-
<!-- Second row: 2 videos -->
403352
<div class="columns is-centered">
404353
<div class="column is-5">
405354
<h4 class="title is-5">Leader-Follower</h4>
@@ -417,7 +366,6 @@ <h4 class="title is-5">Scripted Data Collection</h4>
417366
<br>
418367
<hr>
419368
<h3 class="title is-4 has-text-centered">Reinforcement Learning</h3>
420-
<!-- 2 videos -->
421369
<div class="columns is-centered">
422370
<div class="column is-5">
423371
<video autoplay muted loop preload="metadata" width="75%">
@@ -434,7 +382,6 @@ <h3 class="title is-4 has-text-centered">Reinforcement Learning</h3>
434382
<br>
435383
<hr>
436384
<h3 class="title is-4 has-text-centered">VLA Inference</h3>
437-
<!-- 2 videos -->
438385
<div class="columns is-centered">
439386
<div class="column is-5">
440387
<h4 class="title is-5">Real</h4>
@@ -455,9 +402,6 @@ <h4 class="title is-5">Simulation</h4>
455402
</div>
456403
</div>
457404
</section>
458-
<!-- End Applications -->
459-
460-
<!-- Results -->
461405
<section class="section hero is-small">
462406
<div class="container is-max-desktop">
463407
<div class="columns is-centered has-text-centered">
@@ -467,15 +411,13 @@ <h2 class="title is-3 has-text-centered">Results</h2>
467411
<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.
468412
</p>
469413
<br>
470-
<!-- Left column: Image -->
471414
<div class="columns is-vcentered">
472415
<div class="column is-half">
473416
<figure class="image">
474417
<img src="static/images/results/robot_comparison.svg" alt="Descriptive alt text" width="90%">
475418
</figure>
476419
</div>
477420

478-
<!-- Right column: 2x2 grid of videos -->
479421
<div class="column is-half">
480422
<div class="columns is-multiline">
481423
<div class="column is-half">
@@ -517,17 +459,25 @@ <h2 class="title is-3 has-text-centered">Results</h2>
517459
Success rates in simulation correlate with success rates on the physical robot—consistent with a good evaluation metric.
518460
Adding simulated data to the training mix improves performance in both settings.
519461
</div>
462+
463+
<br>
464+
<hr>
465+
466+
<h3 class="title is-4 has-text-centered">System Performance</h3>
467+
<p>
468+
To support modern, high-frequency VLA architectures (like &pi;<sub>0</sub>), RCS must process data efficiently. We evaluated the recording frequencies across multiple camera and tactile sensor configurations. RCS scales reliably up to 90-120 Hz before encountering hardware IO bottlenecks, proving its suitability for high-frequency RL and VLA deployments despite the synchronous nature of Gymnasium environments.
469+
</p>
470+
<br>
471+
<img src="static/images/system_performance.svg" alt="Configured control vs. measured data frequency during teleoperation" width="100%">
472+
<div class="content has-text-justified">
473+
Fig. 4: Configured control vs. measured data frequency during teleoperation, averaged over more than 1000 steps. The shaded area denotes the standard deviation. <em>FR3 2 Cams</em>: Two RealSense cameras. <em>FR3 4+2 Cams</em>: Four RealSense cameras and two DIGIT sensors. <em>Dual FR3 4+2 Cams</em>: Like <em>FR3 4+2 Cam</em> but with two FR3 robots.
474+
</div>
475+
520476
</div>
521477
</div>
522478
</div>
523479
</div>
524480
</section>
525-
<!-- TODO: explain results -->
526-
<!-- End Results -->
527-
528-
529-
530-
<!--BibTex citation -->
531481
<section class="section" id="BibTeX">
532482
<div class="container is-max-desktop content">
533483
<h2 class="title">BibTeX</h2>
@@ -541,9 +491,6 @@ <h2 class="title">BibTeX</h2>
541491
</code></pre>
542492
</div>
543493
</section>
544-
<!--End BibTex citation -->
545-
546-
547494
<footer class="footer">
548495
<div class="container">
549496
<div class="columns is-centered">
@@ -557,7 +504,6 @@ <h2 class="title">BibTeX</h2>
557504
</div>
558505
</footer>
559506

560-
<!-- Default Statcounter code for Robot Control Stack https://robotcontrolstack.github.io/ -->
561507
<script type="text/javascript">
562508
var sc_project = 13169805;
563509
var sc_invisible = 1;
@@ -570,8 +516,6 @@ <h2 class="title">BibTeX</h2>
570516
src="https://c.statcounter.com/13169805/0/64083f10/1/" alt="free hit counter"
571517
referrerPolicy="no-referrer-when-downgrade"></a></div>
572518
</noscript>
573-
<!-- End of Statcounter Code -->
574-
575-
</body>
519+
</body>
576520

577-
</html>
521+
</html>

0 commit comments

Comments
 (0)