-
Notifications
You must be signed in to change notification settings - Fork 5
/
index.html
224 lines (190 loc) · 10.7 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
---
layout: home
title: Learning Quadrupedal Locomotion over Challenging Terrain
author: Joonho Lee
---
<figure class="figure" id="fig1">
<img alt="compilation-of-anymal-in-the-wild" src="assets/img/figure1.png" class="figure">
</figure>
<div class="paper-title">
{{ page.title }}
</div>
<div class="paper-authors">
<a href="https://scholar.google.com/citations?user=6Htb7swAAAAJ&hl=en">Joonho Lee<sup>1,*</sup></a>,
<a href="https://scholar.google.com/citations?user=Uam1ZB8AAAAJ&hl=en&oi=ao">Jemin Hwangbo<sup>1,2,†</sup></a>,
<a href="https://scholar.google.com/citations?user=0otNjsQAAAAJ&hl=en&oi=ao">Lorenz Wellhausen<sup>1</sup></a>,
<a href="http://vladlen.info/">Vladlen Koltun<sup>3</sup></a>,
and <a href="https://scholar.google.ch/citations?user=DO3quJYAAAAJ&hl=en">Marco Hutter<sup>1</sup></a>
</div>
<div class="paper-affiliations">
<sup>1</sup> Robotic Systems Lab, ETH Zurich
<br><sup>2</sup> Robotics and Artificial Intelligence Lab, KAIST
<br><sup>3</sup> Intelligent Systems Lab, Intel
<br>
<em>
<br><sup>†</sup> Substantial part of the work was carried out during his stay at 1
<br><sup>*</sup> Corresponding author: jolee@ethz.ch
</em>
</div>
<h3>Paper links</h3>
<span style="color:#707070">
<p class="paper-paragraph">
<a href="https://robotics.sciencemag.org/content/5/47/eabc5986">Science Robotics Vol.5 eabc5986 (2020)</a>
<br>Author's version <a href="https://arxiv.org/abs/2010.11251">PDF</a>
</p></span>
<hr class="thick2">
<h2 id="abs">ABSTRACT</h2>
<p class="paper-paragraph">
Legged locomotion can extend the operational domain of robots to some of the most challenging environments
on Earth. However, conventional controllers for legged locomotion are based on elaborate state machines that
explicitly trigger the execution of motion primitives and reflexes. These designs have increased in complexity but
fallen short of the generality and robustness of animal locomotion. Here, we present a robust controller for blind
quadrupedal locomotion in challenging natural environments. Our approach incorporates proprioceptive feedback
in locomotion control and demonstrates zero-shot generalization from simulation to natural environments. The
controller is trained by reinforcement learning in simulation. The controller is driven by a neural network policy
that acts on a stream of proprioceptive signals. The controller retains its robustness under conditions that were
never encountered during training: deformable terrains such as mud and snow, dynamic footholds such as rubble,
and overground impediments such as thick vegetation and gushing water. The presented work indicates that
robust locomotion in natural environments can be achieved by training in simple domains.
</p>
<hr class="thick">
<figcaption>
Summary video
</figcaption>
<div class="videoWrapper" id="movie1">
<iframe width="560" height="315" src="https://www.youtube.com/embed/9j2a1oAHDL8" frameborder="0"
allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture"
allowfullscreen></iframe>
</div>
<div>
<hr class="thick2">
<h2 id="featured">Highlights</h2>
<div style=" padding-bottom: 1rem;">
<em><h3>Zero-shot generalization to unknown environments</h3></em>
</div>
<figcaption>
Figure 1. A number of specific deployments
<span style="font-size: 1rem; color:#595959">(paper Figure 2)</span>
</figcaption>
<figure class="figure" id="fig2">
<img alt="figure2" src="assets/img/figure2.png">
</figure>
<p class="paper-paragraph">
The presented controller has been deployed in diverse natural environments.
These include steep mountain trails, creeks with running water, mud, thick vegetation, loose rubble,
snow-covered
hills, and a damp forest.
A number of specific scenarios are further highlighted in <a href="#fig2">Fig. 1</a>A-F. These environments have
characteristics that the policy does not experience during training. The terrains can deform and crumble, with
significant variation of material properties over the surface. The robot's legs are subjected to frequent
disturbances due to vegetation, rubble, and sticky mud.
Existing terrain estimation pipelines that use cameras or LiDAR fail in environments
with
snow (<a href="#fig2">Fig. 1</a>A), water (<a href="#fig2">Fig. 1</a>C),
or dense vegetation (<a href="#fig2">Fig. 1</a>F).
Our controller does not rely on exteroception and is immune to such failure.
The controller learns omnidirectional locomotion based on a history of proprioceptive observations and
is robust in zero-shot deployment on terrains with characteristics that were never experienced during training.
</p>
<p class="paper-paragraph">
Our controller was used by the Cerberus team for the DARPA Subterranean Challenge Urban Circuit
(<a href="#fig2">Fig. 1</a>G).
It replaced a model-based controller that had been employed used by the team in the past.
The objective of the competition is to develop robotic systems that rapidly map, navigate,
and search complex underground environments, including tunnels, urban underground, and cave
networks. The human operators are not allowed to assist the robots during the competition physically;
only teleoperation is allowed. Accordingly, the locomotion controller needs to perform without failure
over extended mission durations.
The presented controller drove two ANYmal-B robots in four missions of 60 minutes. The
controller exhibited a zero failure rate throughout the competition. A steep staircase that was traversed
by one of the robots during the competition is shown in <a href="#fig2">Fig. 1</a>G.
</p>
<hr class="thick">
<div style=" padding-bottom: 1rem;">
<em><h3>Foot-trapping reflex</h3></em>
</div>
<figcaption>
Movie 1. Step experiment
<span style="font-size: 1rem; color:#595959">(paper Movie S3)</span>
</figcaption>
<div class="videoWrapper" id="movieS3">
<iframe width="560" height="315" src="https://www.youtube.com/embed/tPixnjLbTvE" frameborder="0"
allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture"
allowfullscreen></iframe>
</div>
<p class="paper-paragraph">
The learned controller manifests a foot-trapping reflex, as shown in <a href="#movieS3">Movie 1</a>.
The policy identifies the trapping of the foot purely from proprioceptive observations and lifts the foot over
the obstacle.
Such reflexes were not specified in any way during training: they developed adaptively.
This distinguishes the presented approach from conventional controller design methods,
which explicitly build in such reflexes and orchestrate their execution by a higher-level state machine.
</p>
<hr class="thick">
<div style=" padding-bottom: 1rem;">
<em><h3>Robustness to Model-mismatch</h3></em>
</div>
<figcaption>
Movie 2. Payload experiment
<span style="font-size: 1rem; color:#595959">(paper Movie S4)</span>
</figcaption>
<div class="videoWrapper" id="movieS4">
<iframe width="560" height="315" src="https://www.youtube.com/embed/3Nr47MXCFO0" frameborder="0"
allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture"
allowfullscreen></iframe>
</div>
<p class="paper-paragraph">
We tested the controllers in the presence of substantial model mismatch.
We attached a 10 kg payload.
This payload is 22.7 % of the total weight of the robot, and was never simulated during training.
As shown in <a href="#movieS4">Movie 2</a>, the presented controller can still traverse steps up to 13.4 cm
despite the model mismatch.
The baseline is incapable of traversing any steps under any command speed with the payload.
</p>
<hr class="thick">
<div style=" padding-bottom: 1rem;">
<em><h3>Robustness to Foot Slippage</h3></em>
</div>
<figcaption>
Movie 3. Foot Slippage experiment
<span style="font-size: 1rem; color:#595959">(paper Movie S5)</span>
</figcaption>
<figure class="videoWrapper">
<iframe width="560" height="315" src="https://www.youtube.com/embed/aMPwB3t4idU" frameborder="0"
allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture"
allowfullscreen></iframe>
</figure>
<p class="paper-paragraph">
Next we test robustness to foot slippage.
To introduce slippage, we used a moistened whiteboard.
The results are shown in <a href="#movieS5">Movie 3</a>.
The baseline quickly loses balance, aggressively swings the legs, and falls. In contrast, the presented
controller adapts to the slippery terrain and successfully locomotes in the commanded direction.
</p>
</div>
<p>
</p>
<hr class="thick2">
<div id="acknowlege">
<h2>Acknowledgment</h2>
<p class="paper-paragraph">
<b>Author contributions</b>: J.L. formulated
the main idea of the training and control methods, implemented the controller, set up the
simulation, and trained control policies. J.L. performed the indoor experiments. J.H.
contributed in setting up the simulation. J.L. and L.W. performed outdoor experiments
together. J.L., J.H., L.W., M.H., and V.K. refined ideas, contributed in the experiment design,
and analyzed the data.
</p>
<p class="paper-paragraph">
<b>Funding</b>: The project was funded, in part, by the Intel
Network on Intelligent Systems, the Swiss National Science Foundation (SNF) through the
National Centre of Competence in Research Robotics, the European Research Council (ERC)
under the European Union’s Horizon 2020 research and innovation programme grant
agreement no 852044 and no 780883. The work has been conducted as part of ANYmal
Research, a community to advance legged robotics.
<!-- <b>Competing interests</b>: The authors declare that they have no competing-->
<!-- interests. <b>Data and materials availability</b>: All data needed to evaluate the conclusions in the-->
<!-- paper are present in the paper or the Supplementary Materials. Other materials can be found at-->
<!-- https://github.com/leggedrobotics/learning_locomotion_over_challening_terrain_supplementary.-->
</p>
</div>