-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
315 lines (294 loc) · 15.9 KB
/
index.html
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
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Towards Learning-based Control for Robust Real-world Robotic Grasping in Dynamic Environments</title>
<style>
body {
font-family: Arial, sans-serif;
line-height: 1.6;
color: #333;
max-width: 1200px;
margin: 0 auto;
padding: 20px;
text-align: center;
}
h1, h2 {
color: #2c3e50;
}
.video-grid {
display: grid;
grid-template-columns: repeat(auto-fit, minmax(250px, 1fr));
gap: 20px;
margin-top: 20px;
}
.video-item {
width: 100%;
}
video {
width: 100%;
height: auto;
}
ul {
list-style-position: inside;
padding-left: 0;
}
</style>
<script>
MathJax = {
tex: {
inlineMath: [['$', '$'], ['\\(', '\\)']],
displayMath: [['$$', '$$'], ['\\[', '\\]']],
processEscapes: true
}
};
</script>
<!-- Load MathJax -->
<script id="MathJax-script" async src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script>
</head>
<body>
<header>
<h1>Towards Learning-based Control for Versatile Robotic Grasping in the Real World</h1>
<p><font size="+2"> <em>Authors:</em> Nicolas Bach, Christian Jestel, Julian Eßer, Oliver Urbann and Peter Detzner </font><br>
<font size="+1"> Department of AI and Autonomous Systems, Fraunhofer Institute for Material Flow and Logistics (IML) </font></p>
</header>
<main>
<section>
<h2>Abstract</h2>
<p>Robotic manipulation in non-structured environments presents significant challenges, especially compared to the adaptability and flexibility of humans. While traditional robotic systems excel in controlled settings, their performance falters in unpredictable scenarios. Learning-based control has shown promise in addressing these challenges by developing adaptable behaviors for robotic platforms. However, its application to real-world manipulation tasks remains limited. In this paper we present a two-stage training process that generates versatile and robust policies for robotic grasping tasks in the real-world. In particular, we introduce new rewards and observations of net contact measurements for more effective teacher training. Moreover, we utilize privileged information to inform point cloud sampling, enhancing student training and sim-to-real transfer reliability. Our training process is validated through ablation studies and real-world experiments, demonstrating robust grasping of various objects under a variety of changing environmental conditions. These advancements contribute to bridging the sim-to-real gap, paving the way for generalizable and deployable manipulation policies that function independently of specific settings. </p>
</section>
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/Franka_PaperVideo2.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<section>
<h2>Method</h2>
<figcaption>Overview of the Teacher-Student training process. First, we train a teacher policy on privileged information using Reinforcement Learning. After the policy achieves a sufficient success rate, we use the teacher in an imitation learning
process to generate target actions, that the student imitates. Additionally, the student estimates the current distance between
the grasp position and the object center. We use this estimation for the informed sampling process, which generates an
object-centric point cloud, which we merge with a synthetic point cloud representing the robot.</figcaption>
<figure>
<img src="assets/images/architecture-1.png" alt="Example 5" style="max-width: 100%; height: auto;">
</figure>
<br>
<h3> Uniform Sampling (left) vs. Object Tracking with Auxiliary Head and Informed Sampling (right) </h3>
<br>
Here as a visuell example we show the common way of uniformely sampling and our informed sampling-method from a point cloud. We only use the object estimation output generated by the policy to set the object center and sample the point cloud using our proposed method. The policy is able to track the object on the table.
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/object_tracking_uniform_pc.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/object_tracking_informed_sampling_short.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<br>
</section>
<section>
<h2>Quantitative Experiments</h2>
To evaluate the proposed methods, we perform an ablation study to assess the extension to the baseline in simulation. Further, we evaluate the efficiency of the proposed point cloud sampling method. In real-world experiments, we investigate the trained policies in terms of grasping success and robustness to deviations, such as different scenes, perturbations, and camera positions.
<h3>Grasping Experiments</h3>
The grasping experiments we conducted of all twelve objects.
<br>
\[
\begin{array}{l | c c | c }
\textbf{Object} & \textbf{Ours} & \textbf{Avg. Grasp Time} & \textbf{Wang et al.} \\
\hline
Screwdriver & \textbf{5/5} & 9.00\,s & N/A\\
Can & \textbf{4/5} & 9.75\,s & 3/5\\
Mug & \textbf{5/5} & 8.20\,s & 4/5\\
Banana & 5/5 & 14.20\,s & N/A \\
Brick & 5/5 & 9.60\,s & 5/5\\
Soup Can & 3/5 & 15.70\,s & 3/5\\
Sugar Box & \textbf{5/5} & 8.60\,s & 4/5\\
Cracker Box & 2/5 & 17.50\,s & \textbf{3/5}\\
Mustard & 4/5 & 12.50\,s & 4/5\\
Ball & 4/5 & 22.25\,s & N/A\\
Bowl & \textbf{5/5} & 9.80\,s & 4/5\\
Bleacher & 4/5 & 13.50\,s & 4/5\\
\hline
In Comparison & \textbf{37/45} & - & 34/45\\
Success Rate & \textbf{82.2%} & - & 75.6\,\%\\
\hline
All & 51/60 & 12.54\,s & -\\
Success Rate & 85.0% & - & - \\
\end{array}
\]
<br>
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_00_screwdriver.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_01_meat_can.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_02_mug.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_03_banana.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_04_brick.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_05_soup_can.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_06_sugar_box.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_07_cracker_box.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_08_mustard.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_09_ball.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_10_bowl.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/experiment_11_bleacher.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<br>
</section>
<section>
<h2> Qualitative Experiments </h2>
<h3>Invariance to Changes in the Scene</h3>
In this experiment we perform two grasps of a mug, then change the camera and move the surface, from which we grasp the object.
Then we perform two more grasp and change the camera and desk again, to perform two final grasps. This shows, that our method works without any strong synchronization of simulated scene and reality.
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/invariance_to_scene_changes.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<br>
<h3>Different Scene and Camera Angle</h3>
We turn the desk, that the robot is situated on towards another surface, that is strongly out-of-distribution from the environment, that we trained the policy in. Furthermore, the camera angle and the robot pose is also not included in simulation training. However, by leveraging informed point cloud sampling and the sim-to-real methods we propose, the policy still has some success in grasping objects.
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/drastic_different_scene_mug.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/drastic_different_scene_sugar_box_04.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/drastic_different_scene_meat_can_02.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<br>
<h3>Grasping with Distractors</h3>
We test the method on performing in scenes with different objects. If there is a big difference between the object to be grasped and the distractors, the policy can easily choose to grasp the right object thanks to informed sampling.
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/multi_object_scene_banana.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/multi_object_scene_sugar_box.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<br>
<h3>Failures</h3>
Here, we depict some failures of the system. To be precise, grasps where the system took especially long to perform the task or where a human had to intervene. The most common failure case is slightly failing in the grasp due to precision and then repeating this behavior over and over again. Sometimes the system also fails in estimating the state correctly resulting in an orientation that doesn't allow a successful grasp.
<div class="video-grid">
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/looping_behavior_during_experiments.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/failures.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video controls>
<source src="assets/videos/real_world_experiments/fail_grasp_banana_comp.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
</section>
</main>
<footer>
<a href="https://www.iml.fraunhofer.de/en/imprint.html" class="link">Imprint</a>
<a href="https://www.iml.fraunhofer.de/en/data_protection.html" class="link">Data Protection Policy</a>
</footer>
</body>
</html>