-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathresearch.html
782 lines (722 loc) · 41.8 KB
/
research.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
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
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=900">
<title>Research</title>
<link rel="stylesheet" href="styles.css">
<link rel="icon" href="assets/icon.jpg">
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Castoro"
rel="stylesheet">
<link href="https://fonts.googleapis.com/css?family=Noto+Sans:400,700,400italic,700italic"
rel="stylesheet">
<!-- <link href='http://fonts.googleapis.com/css?family=Lato:400,700,400italic,700italic' rel='stylesheet'> -->
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.15.3/css/all.min.css">
</head>
<body>
<table width="900" border="0" align="center" cellspacing="0" cellpadding="20">
<tr>
<td style="width:25%; vertical-align:middle; padding-right: 10px;">
<a href="index.html"><img src="assets/logo.png" height="100"></a>
</td>
<td style="width:15%; vertical-align:middle; text-align:center; padding-right: 5px;">
<a href="people.html" style="font-size: 22px; color:black">People</a>
</td>
<td style="width:15%; vertical-align:middle; text-align:center; padding-right: 10px;">
<a href="publications.html" style="font-size: 22px; color:black">Publications</a>
</td>
<td style="width:15%; vertical-align:middle; text-align:center; padding-right: 0px;">
<a href="research.html" style="font-size: 22px; color:black">Research</a>
</td>
<td style="width:15%; vertical-align:middle; text-align:center; padding-right: 0px;">
<a href="robots.html" style="font-size: 22px; color:black">Robots</a>
</td>
<td style="width:15%; vertical-align:middle; text-align:center; padding-right: 0px;">
<a href="join.html" style="font-size: 22px; color:black">Join</a>
</td>
</tr>
</table>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Research Talks</heading>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<ul>
<li>2024 Oct: CMU RI Seminar Talk "Building Generalist Robots with Agility via Learning and Control: Humanoids and Beyond" (one hour) [<a href="https://youtu.be/Uym3Tr6t5TM?si=dJWQgO-mxbUlIwoQ" target="_blank"><i class="fas fa-globe"></i> recording</a>] [<a href="https://youtu.be/pVTG7809RTg?si=6BnsMqveGO_KIrdn" target="_blank"><i class="fas fa-globe"></i> 2025 April version at ETH Zürich</a>] [<a href="https://drive.google.com/file/d/1iyMGxjK5ID-j-ga3B7sUsqN3erle1ibY/view?usp=sharing" target="_blank"><i class="far fa-file"></i> slides</a>]</li>
<li>2024 Sep: Georgia Tech IRIM Seminar Talk "Unifying Semantic and Physical Intelligence for Generalist Humanoid Robots" (one hour) [<a href="https://mediaspace.gatech.edu/media/1_tckra9zz" target="_blank"><i class="fas fa-globe"></i> recording</a>]</li>
<!-- <li>2023 Sep: New faculty lighting talk at CMU SCS (5 mins) [<a href="https://youtu.be/1MH-R6_UALw?si=isH5XlHQzg1719jy" target="_blank"><i class="fas fa-globe"></i> recording</a>]</li> -->
</ul>
</td>
</tr>
</table>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Learning and Control for Humanoid Locomotion and Loco-Manipulation</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='assets/OmniH2O.gif' width="35%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Humanoid robots offer two unparalleled advantages in general-purpose embodied intelligence. First, humanoids are built as generalist robots that can potentially do all the tasks humans can do in complex environments. Second, the embodiment alignment between humans and humanoids allows for the seamless integration of human cognitive skills with versatile humanoid capabilities, making humanoids the most promising physical embodiment for AI.
</p>
<p>Humanoid control in a whole-body manner is challenging, due to its high degrees of freedom and contact-rich nature (see this <a href="https://arxiv.org/abs/2501.02116" target="_blank">survey</a>). We aim to develop learning-based whole-body control methods for humanoid locomotion and loco-manipulation problems, enabling humanoids to interact with the physical world, adapt quickly to many tasks and environments, and integrate with high-level decision-making layers such as VLMs.
</p>
</td>
</tr>
</table>
<details style="width: 880px; margin: 0 auto;">
<summary style="cursor: pointer; padding: 0px 0;">
<papertitle>Selected papers in this topic:</papertitle>
</summary>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='assets/ASAP.gif' width="95%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills</papertitle>
<br>
Tairan He<sup>*</sup>, Jiawei Gao<sup>*</sup>, Wenli Xiao<sup>*</sup>, Yuanhang Zhang<sup>*</sup>, Zi Wang, Jiashun Wang, Zhengyi Luo, Guanqi He, Nikhil Sobanbab, Chaoyi Pan, Zeji Yi, Guannan Qu, Kris Kitani, Jessica Hodgins, Linxi "Jim" Fan, Yuke Zhu, Changliu Liu, Guanya Shi
<br>
<a href="https://arxiv.org/abs/2502.01143" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://agile.human2humanoid.com/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/LeCAR-Lab/ASAP" target="_blank"><i class="fas fa-code"></i> code</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: ASAP learns agile whole-body humanoid motions via learning a residual action model from the real world to align sim and real physics.
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr style="background-color: var(--highlight-color)">
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='assets/OmniH2O.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>OmniH2O: Universal and Dexterous Human-to-Humanoid Whole-Body Teleoperation and Learning</papertitle>
<br>
Tairan He<sup>*</sup>, Zhengyi Luo<sup>*</sup>, Xialin He<sup>*</sup>, Wenli Xiao, Chong Zhang, Weinan Zhang, Kris Kitani, Changliu Liu, Guanya Shi
<br>
<em>Conference on Robot Learning (CoRL)</em>, 2024
<br>
<a href="https://arxiv.org/abs/2406.08858" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://omni.human2humanoid.com/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://cmu.box.com/s/kmayzq5ax2rxvwn97s0hzz0aq5vws9io" target="_blank"><i class="fas fa-database"></i> dataset</a>  
<a href="https://github.com/LeCAR-Lab/human2humanoid" target="_blank"><i class="fas fa-code"></i> code</a>  
<a href="https://spectrum.ieee.org/video-friday-drone-vs-flying-canoe" target="_blank"><i class="fas fa-newspaper"></i> IEEE Spectrum</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: OmniH2O provides a universal whole-body humanoid control interface that enables diverse teleoperation and autonomy methods.
</p>
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='assets/WoCoCo.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>WoCoCo: Learning Whole-Body Humanoid Control with Sequential Contacts</papertitle>
<br>
Chong Zhang<sup>*</sup>, Wenli Xiao<sup>*</sup>, Tairan He, Guanya Shi
<br>
<em>Conference on Robot Learning (CoRL)</em>, 2024
<p style="color: orange; margin: 0px 0;">(Oral Presentation)</p>
<a href="https://arxiv.org/abs/2406.06005" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://lecar-lab.github.io/wococo/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://spectrum.ieee.org/video-friday-drone-vs-flying-canoe" target="_blank"><i class="fas fa-newspaper"></i> IEEE Spectrum</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: WoCoCo is a task-agnostic skill learning framework without any motion priors, by decomposing long-horizon tasks into contact sequences.
</p>
</td>
</tr>
</table>
</details>
<br>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Improve Offline Learned Policies via Online Adaptation</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='research/adaptive.png' width="35%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Offline learned policies for robotic control have shown great success in the robot learning community. For example, learning manipulation policies from demonstrations using imitation learning with generative models; learning locomotion policies in simulation using sim2real reinforcement learning. However, those policies are frozen in test time, and cannot efficiently adapt to new environments or tasks. We aim to develop methods that can effectively learn "adaptable" representation from offline data and efficiently adapt in real time.
</p>
</td>
</tr>
</table>
<details style="width: 880px; margin: 0 auto;">
<summary style="cursor: pointer; padding: 0px 0;">
<papertitle>Selected papers in this topic:</papertitle>
</summary>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<img src='publications/2022_neural_fly.gif' width="90%">
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Neural-Fly Enables Rapid Learning for Agile Flight in Strong Winds</papertitle>
<br>
Michael O'Connell<sup>*</sup>, Guanya Shi<sup>*</sup>, Xichen Shi, Kamyar Azizzadenesheli, Animashree Anandkumar, Yisong Yue, Soon-Jo Chung
<br>
<em>Science Robotics</em>
<br>
<a href="https://arxiv.org/abs/2205.06908" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://youtu.be/TuF9teCZX0U" target="_blank"><i class="fas fa-video"></i> video</a>  
<a href="https://www.caltech.edu/about/news/rapid-adaptation-of-deep-learning-teaches-drones-to-survive-any-weather" target="_blank"><i class="fas fa-newspaper"></i> Caltech news</a>  
<a href="https://youtu.be/R1S5BnKgJxs" target="_blank"><i class="fas fa-newspaper"></i> Reuters</a>  
<a href="https://www.cnn.com/videos/business/2022/05/31/caltech-neural-fly-drones-in-strong-wind-orig-ht.cnn-business/video/playlists/business-tech/" target="_blank"><i class="fas fa-newspaper"></i> CNN</a>  
<a href="https://github.com/aerorobotics/neural-fly" target="_blank"><i class="fas fa-code"></i> code</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: Neural-Fly uses adaptive control to online fine-tune a meta-pretrained DNN representation, enabling rapid adaptation in strong winds.
</p>
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2023_datt.gif' width="62%">
<img src='publications/2023_datt.png' width="38%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>DATT: Deep Adaptive Trajectory Tracking for Quadrotor Control</papertitle>
<br>
Kevin Huang, Rwik Rana, Alexander Spitzer, Guanya Shi, Byron Boots
<br>
<em>Conference on Robot Learning (CoRL)</em>, 2023
<p style="color: orange; margin: 0px 0;">(Oral presentation, 6.6%)</p>
<a href="https://openreview.net/pdf?id=XEw-cnNsr6" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://sites.google.com/view/deep-adaptive-traj-tracking" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/KevinHuang8/DATT" target="_blank"><i class="fas fa-code"></i> code</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: DATT can precisely track arbitrary, potentially infeasible trajectories in the presence of large disturbances.
</p>
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2024_AnyCar.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>AnyCar to Anywhere: Learning Universal Dynamics Model for Agile and Adaptive Mobility</papertitle>
<br>
Wenli Xiao<sup>*</sup>, Haoru Xue<sup>*</sup>, Tony Tao, Dvij Kalaria, John M. Dolan, Guanya Shi
<br>
<em>Intertional Conference on Robotics and Automation (ICRA)</em>, 2025
<br>
<a href="https://arxiv.org/abs/2409.15783" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://lecar-lab.github.io/anycar/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/LeCAR-Lab/anycar" target="_blank"><i class="fas fa-code"></i> code</a>  
<a href="https://spectrum.ieee.org/video-friday-mobile-robot-upgrades" target="_blank"><i class="fas fa-newspaper"></i> IEEE Spectrum</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: AnyCar is a transformer-based dynamics model that can adapt to various vehicles, environments, state estimators, and tasks.
</p>
</td>
</tr>
</table>
</details>
<br>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Real2Sim2Real Reinforcement Learning</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='assets/ASAP.gif' width="35%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Sim2real Reinforcement Learning (RL) has become the dominant method for many locomotion problems, especially for humanoids, legged robots, drones, and ground vehicles. However, the large sim2real gap, tedious reward tuning, and the lack of diversity hinder its applications in other problems such as loco-manipulation and dexterous manipulation. We aim to address these limitations via enhancing the simulation training environment using real-world data, which is often referred to as <em>real2sim2real</em>.
</p>
</td>
</tr>
</table>
<details style="width: 880px; margin: 0 auto;">
<summary style="cursor: pointer; padding: 0px 0;">
<papertitle>Selected papers in this topic:</papertitle>
</summary>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr style="background-color: var(--highlight-color)">
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='assets/ASAP.gif' width="95%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills</papertitle>
<br>
Tairan He<sup>*</sup>, Jiawei Gao<sup>*</sup>, Wenli Xiao<sup>*</sup>, Yuanhang Zhang<sup>*</sup>, Zi Wang, Jiashun Wang, Zhengyi Luo, Guanqi He, Nikhil Sobanbab, Chaoyi Pan, Zeji Yi, Guannan Qu, Kris Kitani, Jessica Hodgins, Linxi "Jim" Fan, Yuke Zhu, Changliu Liu, Guanya Shi
<br>
<a href="https://arxiv.org/abs/2502.01143" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://agile.human2humanoid.com/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/LeCAR-Lab/ASAP" target="_blank"><i class="fas fa-code"></i> code</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: ASAP learns agile whole-body humanoid motions via learning a residual action model from the real world to align sim and real physics.
</td>
</tr>
</table>
</details>
<br>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Model-Based RL, World Model, and Sampling-Based Optimal Control</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='publications/2024_DIAL-MPC.gif' width="35%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Model-based RL (MBRL) first learns a dynamics model (or a "world" model) and then generates a policy via planning/optimization, policy learning, or search. The most compelling part of MBRL is that the learned model is <em>task-agnostic</em>. We are interested in all aspects of MBRL, including model learning, theoretical framework, and planning using optimal control. In particular, recently, we have developed new sampling-based optimal control frameworks that enable efficient online decision-making.
</p>
</td>
</tr>
</table>
<details style="width: 880px; margin: 0 auto;">
<summary style="cursor: pointer; padding: 0px 0;">
<papertitle>Selected papers in this topic:</papertitle>
</summary>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2024_DIAL-MPC.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Full-Order Sampling-Based MPC for Torque-Level Locomotion Control via Diffusion-Style Annealing</papertitle>
<br>
Haoru Xue<sup>*</sup>, Chaoyi Pan<sup>*</sup>, Zeji Yi, Guannan Qu, Guanya Shi
<br>
<em>Intertional Conference on Robotics and Automation (ICRA)</em>, 2025
<br>
<a href="https://arxiv.org/abs/2409.15610" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://lecar-lab.github.io/dial-mpc/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/LeCAR-Lab/dial-mpc" target="_blank"><i class="fas fa-code"></i> code</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: DIAL-MPC is the first training-free method achieving real-time whole-body torque control using full-order dynamics.
</p>
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2025_TDMPC_Square.gif' width="95%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>TD-M(PC)<sup>2</sup>: Improving Temporal Difference MPC Through Policy Constraint</papertitle>
<br>
Haotian Lin, Pengcheng Wang, Jeff Schneider, Guanya Shi
<br>
<a href="https://arxiv.org/abs/2502.03550v1" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://darthutopian.github.io/tdmpc_square/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/DarthUtopian/tdmpc_square_public" target="_blank"><i class="fas fa-code"></i> code</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: We observe the value overestimation issue in planner-based MBRL and propose a policy constraint solution with SOTA performance.
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2023_optimal_exploration.png' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Optimal Exploration for Model-based RL in Nonlinear Systems</papertitle>
<br>
Andrew Wagenmaker, Guanya Shi, Kevin Jamieson
<br>
<em>Neural Information Processing Systems (NeurIPS)</em>, 2023
<p style="color: orange; margin: 0px 0;">(Spotlight, 3.1%)</p>
<a href="https://arxiv.org/abs/2306.09210" target="_blank"><i class="far fa-file"></i> paper</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: Not all model parameters are equally important. We develop an instance-optimal exploration algorithm for MBRL in nonlinear systems.
</p>
</td>
</tr>
</table>
</details>
<br>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Algorithmic and Theoretical Foundations for “Computational Control”</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='publications/2024_MBD.gif' width="35%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Compared to learning-based methods, one problem of traditional control methods is that their performance cannot effectively scale up with the amount of data and parallel computing. Our focus is around a new concept called <em>computational control</em>, i.e., control methods that can effectively scale up with the amount of data and parallel computing. Examples include sampling-based optimal control and "generative action model" (policy learning using generative models). The goal is to systematically understand, analyze, and enhance computational control methods and apply them to robotics. One of recent interests is to understand how diffusion/flow-based methods work in decision-making problems.
</p>
</td>
</tr>
</table>
<details style="width: 880px; margin: 0 auto;">
<summary style="cursor: pointer; padding: 0px 0;">
<papertitle>Selected papers in this topic:</papertitle>
</summary>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2024_MBD.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Model-Based Diffusion for Trajectory Optimization</papertitle>
<br>
Chaoyi Pan<sup>*</sup>, Zeji Yi<sup>*</sup>, Guanya Shi<sup>†</sup>, Guannan Qu<sup>†</sup>
<br>
<em>Neural Information Processing Systems (NeurIPS)</em>, 2024
<br>
<a href="https://drive.google.com/file/d/1kPjD79Cfr9spWulWNVFMRHqTE-mjbGAp/view?usp=sharing" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://lecar-lab.github.io/mbd/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/LeCAR-Lab/model-based-diffusion" target="_blank"><i class="fas fa-code"></i> code</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: MBD is a diffusion-based traj optimization method that directly computes the score function using models without any external data.
</p>
</td>
</tr>
</table>
</details>
<br>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Embodied Intelligence in the Air: General-purpose Aerial Manipulation</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='publications/2025_Flying_Hand.gif' width="35%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>To some extent, existing works on aerial manipulation primarily focus on the aerial perspective, rather than the general-purpose manipulation perspective. The goal is to study aerial manipulation from the embodied intelligence perspective, building general hardware platforms and control methods. Similar to humanoids, we aim to solve the whole-body control problem for aerial manipulation. We are also interested in designing the high-level policy via vision-language-action (VLA) models or imitation learning.
</p>
</td>
</tr>
</table>
<details style="width: 880px; margin: 0 auto;">
<summary style="cursor: pointer; padding: 0px 0;">
<papertitle>Selected papers in this topic:</papertitle>
</summary>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr style="background-color: var(--highlight-color)">
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2025_Flying_Hand.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Flying Hand: End-Effector-Centric Framework for Versatile Aerial Manipulation Teleoperation and Policy Learning</papertitle>
<br>
Guanqi He<sup>*</sup>, Xiaofeng Guo<sup>*</sup>, Luyi Tang, Yuanhang Zhang, Mohammadreza Mousaei, Jiahe Xu, Junyi Geng, Sebastian Scherer, Guanya Shi
<br>
<em>Robotics: Science and Systems (RSS)</em>, 2025
<br>
<a href="https://xiaofeng-guo.github.io/flying_hand.io/static/pdf/flying_hand.pdf" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://xiaofeng-guo.github.io/flying_hand.io/" target="_blank"><i class="fas fa-globe"></i> website</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: A general-purpose aerial manipulation framework with an EE-centric interface that bridges whole-body control and policy learning.
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2024_flying_calligrapher.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Flying Calligrapher: Contact-Aware Motion and Force Planning and Control for Aerial Manipulation</papertitle>
<br>
Xiaofeng Guo<sup>*</sup>, Guanqi He<sup>*</sup>, Jiahe Xu, Mohammadreza Mousaei, Junyi Geng, Sebastian Scherer, Guanya Shi
<br>
<em>IEEE Robotics and Automation Letters (RA-L)</em>, 2024
<br>
<a href="https://arxiv.org/abs/2407.05587" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://xiaofeng-guo.github.io/flying-calligrapher/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://spectrum.ieee.org/video-friday-unitree-talks-robots" target="_blank"><i class="fas fa-newspaper"></i> IEEE Spectrum</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: Flying calligrapher enables precise hybrid motion and contact force control for an aerial manipulator in various drawing tasks.
</p>
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2024_aerial_interaction.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Aerial Interaction with Tactile Sensing</papertitle>
<br>
Xiaofeng Guo, Guanqi He, Mohammadreza Mousaei, Junyi Geng, Guanya Shi, Sebastian Scherer
<br>
<em>International Conference on Robotics and Automation (ICRA)</em>, 2024
<br>
<a href="https://arxiv.org/abs/2310.00142" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://sites.google.com/view/aerial-system-gelsight" target="_blank"><i class="fas fa-globe"></i> website</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: We introduce a new aerial manipulation system that leverages tactile feedback for accurate contact force control and texture detection.
</p>
</td>
</tr>
</table>
</details>
<br>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Structured Reinforcement Learning and Control with Guarantees</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='research/structured.png' width="35%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Most RL algorithms are general for all tasks. In contrast, drastically different control methods are developed for different systems/tasks, and their successes highly rely on structures inside these systems/tasks. We seek to encode these structures and algorithmic principles into black-box RL algorithms, to make RL algorithms more data-efficient, robust, interpretable, and safe. Examples include hierarchical RL and optimal control methods, learning safety filter for RL policies, and learning-based nonlinear control with stability guarantees.
</p>
</td>
</tr>
</table>
<details style="width: 880px; margin: 0 auto;">
<summary style="cursor: pointer; padding: 0px 0;">
<papertitle>Selected papers in this topic:</papertitle>
</summary>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='assets/abs.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion</papertitle>
<br>
Tairan He<sup>*</sup>, Chong Zhang<sup>*</sup>, Wenli Xiao, Guanqi He, Changliu Liu, Guanya Shi
<br>
<em>Robotics: Science and Systems (RSS)</em>, 2024
<p style="color: orange; margin: 0px 0;">(Outstanding Student Paper Award Finalist)</p>
<a href="https://arxiv.org/abs/2401.17583" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://agile-but-safe.github.io/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/LeCAR-Lab/ABS" target="_blank"><i class="fas fa-code"></i> code</a>  
<a href="https://spectrum.ieee.org/video-friday-agile-but-safe" target="_blank"><i class="fas fa-newspaper"></i> IEEE Spectrum</a>  
<a href="https://www.ri.cmu.edu/collision-free-high-speed-robots/" target="_blank"><i class="fas fa-newspaper"></i> CMU News</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: ABS enables fully onboard, agile (>3m/s), and collision-free locomotion for quadrupedal robots in cluttered environments.
</p>
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2024_JumpingCoD.gif' width="90%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Agile Continuous Jumping in Discontinuous Terrains</papertitle>
<br>
Yuxiang Yang, Guanya Shi, Changyi Lin, Xiangyun Meng, Rosario Scalise, Mateo Guaman Castro, Wenhao Yu, Tingnan Zhang, Ding Zhao, Jie Tan, Byron Boots
<br>
<em>Intertional Conference on Robotics and Automation (ICRA)</em>, 2025
<br>
<a href="https://arxiv.org/abs/2409.10923" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://yxyang.github.io/jumping_cod/" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/yxyang/jumping_cod" target="_blank"><i class="fas fa-code"></i> code</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: Continuous, agile, and autonomous quadrupedal jumping via hierarchical model-free RL and model-based control.
</p>
</td>
</tr>
</table>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td style="width:35%; vertical-align:middle; padding-right: 20px;">
<div class="image-container">
<img src='publications/2024_SSML-AC.png' width="62%">
</div>
</td>
<td style="width:65%; vertical-align:middle">
<papertitle>Self-Supervised Meta-Learning for All-Layer DNN-Based Adaptive Control with Stability Guarantees</papertitle>
<br>
Guanqi He, Yogita Choudhary, Guanya Shi
<br>
<em>Intertional Conference on Robotics and Automation (ICRA)</em>, 2025
<br>
<a href="https://arxiv.org/abs/2410.07575" target="_blank"><i class="far fa-file"></i> paper</a>  
<a href="https://sites.google.com/view/ssml-ac-project" target="_blank"><i class="fas fa-globe"></i> website</a>  
<a href="https://github.com/LeCAR-Lab/SSML-AC/tree/main" target="_blank"><i class="fas fa-code"></i> code</a>
<p style="margin-top: 5px"><i class="fas fa-comment-dots"></i> TL;DR: Pretrain a residual dynamics DNN using meta-learning and fine-tune the whole DNN online using adaptive control with stability guarantees.
</p>
</td>
</tr>
</table>
</details>
<br>
<br>
<br>
</body>
</html>
<!-- <table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Safe Learning-based Nonlinear Control with Learned Robotic Agility</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='research/safe_learning_control.png' width="65%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Recent advances in machine learning beckon to applications in autonomous systems. However, for safety-critical settings, the learning system must interact with the rest of the autonomous system in a way that safeguards against catastrophic failures with guarantees. In addition, from computational and statistical standpoints, the learning system must incorporate prior knowledge for efficiency and generalizability. Leveraging control-theoretic tools and prior knowledge, we aim to develop learning-based control methods with both guarantees and new capabilities.
</p>
</td>
</tr>
</table>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Offline Learning and Online Adaptation</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='research/adaptive.png' width="40%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Real-world robotic systems have to operate in unknown and dynamic environments where the decision-maker must fast adapt to uncertainties. For example, legged robot rescue and search necessitates traversing complicated terrain conditions. Deep learning has representation power but is often too slow to update onboard. On the other hand, adaptive control can update as fast as the feedback control loop with guarantees. Our goal is to develop offline and online algorithms that can effectively learn from offline data and efficiently fine-tune/adapt in real time.
</p>
</td>
</tr>
</table>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Structured Reinforcement Learning and Control</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='research/structured.png' width="50%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Most RL algorithms (e.g., TRPO, SAC) are general for all tasks. In contrast, drastically different control methods are developed for different systems/tasks, and their successes highly rely on structures inside these systems/tasks. We seek to encode these structures and algorithmic principles into black-box RL algorithms, to make RL algorithms more data-efficient, robust, interpretable, and safe. We are particularly interested in hierarchical and superpositional RL and control approaches.
</p>
</td>
</tr>
</table>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Learning and Control Theory: Towards a Unified Framework</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='research/theory.png' width="70%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>There are many closely related concepts in machine learning and control communities, for instance, model-based RL and optimal control, online learning and adaptive control, domain randomization and robust control, online optimization and MPC, just to name a few. We seek to build interfaces and unified frameworks, which not only deepen fundamental connections between learning and control, but inspire new algorithms. One example of such connections is analyzing MPC's dynamic regret (a learning-theoretic metric).
</p>
</td>
</tr>
</table>
<br>
<table width="880" border="0" align="center" cellspacing="0" cellpadding="0">
<tr>
<td width="100%" valign="middle">
<heading>Swarm Intelligence</heading>
</td>
</tr>
<tr>
<td style="width:100%; vertical-align:middle">
<div class="image-container">
<img src='research/swarm.png' width="35%">
</div>
</td>
</tr>
<tr>
<td width="100%" valign="middle">
<p>Robot swarm learning and control present multiple new challenges, such as complex interactions between agents and dynamic topology. We aim to develop scalable and robust decision-making methods for multi-agent robotic systems, by leveraging properties like symmetry, locality, and invariance. We are also interested in how different types of robots interact (e.g., drone and legged robot), and human-robot teaming.
</p>
</td>
</tr>
</table>
-->