1<!DOCTYPE html> 2<html lang="en"> 3 <!-- Head --> 4 <head> 5<meta http-equiv="Content-Type" content="text/html; charset=UTF-8"> 6 7 <!-- Metadata, OpenGraph and Schema.org --> 8 9 10 11 12<!-- Standard metadata --> 13<meta charset="utf-8"> 14<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no"> 15<meta http-equiv="X-UA-Compatible" content="IE=edge"> 16<title> 17 18 19 Yuhong Cao 20 21</title> 22<meta name="author" content="Yuhong Cao"> 23<meta name="description" content="A simple, whitespace theme for academics. Based on [*folio](https://github.com/bogoli/-folio) design. 24"> 25 26 <meta name="keywords" content="deep reinforcement learning, robot learning, view planning, path planning"> 27 28 29 30 31 32 33 34 35 36 37<!-- Bootstrap & MDB --> 38<link rel="stylesheet" href="/assets/css/bootstrap.min.css?a4b3f509e79c54a512b890d73235ef04"> 39<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/css/mdb.min.css" integrity="sha256-jpjYvU3G3N6nrrBwXJoVEYI/0zw8htfFnhT9ljN3JJw=" crossorigin="anonymous"> 40 41<!-- Bootstrap Table --> 42 43 44<!-- Fonts & Icons --> 45<link defer rel="stylesheet" href="/assets/css/academicons.min.css?f0b7046b84e425c55f3463ac249818f5"> 46<link defer rel="stylesheet" type="text/css" href="https://fonts.googleapis.com/css?family=Roboto:300,400,500,700|Roboto+Slab:100,300,400,500,700|Material+Icons&display=swap"> 47 48<!-- Code Syntax Highlighting --> 49<link defer rel="stylesheet" href="/assets/css/jekyll-pygments-themes-github.css?591dab5a4e56573bf4ef7fd332894c99" media="" id="highlight_theme_light"> 50 51 52 53<!-- Styles --> 54 55<!-- pseudocode --> 56 57 58 59 <link rel="shortcut icon" href="/assets/img/logo.svg?7b8f72542d151aa3a3fff62f21cbd55d"> 60 61<link rel="stylesheet" href="/assets/css/main.css?d41d8cd98f00b204e9800998ecf8427e"> 62<link rel="canonical" href="https://www.yuhongcao.online/"> 63 64<!-- Dark Mode -->
65<script src="/assets/js/theme.js?9a0c749ec5240d9cda97bc72359a72c0"></script>
65 66 67 <link defer rel="stylesheet" href="/assets/css/jekyll-pygments-themes-native.css?5847e5ed4a4568527aa6cfab446049ca" media="none" id="highlight_theme_dark"> 68
68<script> 69 initTheme(); 70 </script>
70 71 72 73<!-- GeoJSON support via Leaflet --> 74 75 76<!-- diff2html --> 77 78 79 80 81 82 83 </head> 84 85 <!-- Body --> 86 <body class="fixed-top-nav "> 87 <!-- Header --> 88 <header> 89 <!-- Nav Bar --> 90 <nav id="navbar" class="navbar navbar-light navbar-expand-sm fixed-top" role="navigation"> 91 <div class="container"> 92 93 <!-- Navbar Toggle --> 94 <button class="navbar-toggler collapsed ml-auto" type="button" data-toggle="collapse" data-target="#navbarNav" aria-controls="navbarNav" aria-expanded="false" aria-label="Toggle navigation"> 95 <span class="sr-only">Toggle navigation</span> 96 <span class="icon-bar top-bar"></span> 97 <span class="icon-bar middle-bar"></span> 98 <span class="icon-bar bottom-bar"></span> 99 </button> 100 101 <div class="collapse navbar-collapse text-right" id="navbarNav"> 102 <ul class="navbar-nav ml-auto flex-nowrap"> 103 104 105 <!-- About --> 106 <li class="nav-item active"> 107 <a class="nav-link" href="/">About 108 109 <span class="sr-only">(current)</span> 110 111 </a> 112 </li> 113 114 <!-- Other pages --> 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 <li class="nav-item "> 142 143 <a class="nav-link" href="/publications/">Publications 144 145 </a> 146 </li> 147 148 149 150 151 152 153 154 155 156 <li class="nav-item "> 157 158 <a class="nav-link" href="/cv/">CV 159 160 </a> 161 </li> 162 163 164 165 166 167 168 169 170 171 172 173 <!-- Toogle theme mode --> 174 <li class="toggle-container"> 175 <button id="light-toggle" title="Change theme"> 176 <i class="ti ti-sun-moon" id="light-toggle-system"></i> 177 <i class="ti ti-moon-filled" id="light-toggle-dark"></i> 178 <i class="ti ti-sun-filled" id="light-toggle-light"></i> 179 </button> 180 </li> 181 182 </ul> 183 </div> 184 </div> 185 </nav> 186 187 <!-- Scrolling Progress Bar --> 188 <progress id="progress" value="0"> 189 <div class="progress-container"> 190 <span class="progress-bar"></span> 191 </div> 192 </progress> 193 194</header> 195 196 197 <!-- Content --> 198 <div class="container mt-5" role="main"> 199 200 <div class="post"> 201 <header class="post-header"> 202 <h1 class="post-title"> 203 204 Yuhong Cao 205 206 </h1> 207 <p class="desc">Postdoc @ NUS ME</p> 208 </header> 209 210 <article> 211 212 <div class="profile float-right"> 213 214 215 216 217 218 219<figure> 220 <picture> 221 <!-- Auto scaling with imagemagick --> 222 <!-- 223 See https://www.debugbear.com/blog/responsive-images#w-descriptors-and-the-sizes-attribute and 224 https://developer.mozilla.org/en-US/docs/Learn/HTML/Multimedia_and_embedding/Responsive_images for info on defining 'sizes' for responsive images 225 --> 226 227 <source class="responsive-img-srcset" srcset="/assets/img/prof_pic-480.webp 480w,/assets/img/prof_pic-800.webp 800w,/assets/img/prof_pic-1400.webp 1400w," sizes="(min-width: 930px) 270.0px, (min-width: 576px) 228 30vw, 95vw" type="image/webp"> 229 230 <img src="/assets/img/prof_pic.jpg?e553389ba65e66008d9f3a62493d578f" class="img-fluid z-depth-1 231 rounded" width="100%" height="auto" alt="prof_pic.jpg" loading="eager" onerror="this.onerror=null; $('.responsive-img-srcset').remove();"> 232 </source></picture> 233 234 235</figure> 236 237 238 239 <div class="more-info"> 240<!-- <p>Singapore</p> --> 241</div> 242 243 </div> 244 245 246 <div class="clearfix"> 247<p>I am a postdoc fellow at National University of Singapore, Department of Mechanical Engineering, <a href="https://www.marmotlab.org/" rel="external nofollow noopener" target="_blank">MARMot Lab</a>. I did my PhD in NUS under the supervision of <a href="https://www.marmotlab.org/bio.html" rel="external nofollow noopener" target="_blank">Dr. Guillaume Sartoretti</a>. Before that, I got my bachelor degree in Beihang Univeristy in China.</p> 248 249<p>My research began with deep reinforcement learning-based navigation and is evolving toward unifying high-level decision-making with advanced mobility. I aim to develop generalizable learning frameworks with solid systems engineering, to enable intelligent and reliable robot operations in the real world.</p> 250 251<p>In my spare time, I am a casual athlete in basketball and swimming. During my undergraduate, I was the captain of the department basketball team and the coach of the university swimming assiociation (However I found I spend less time on sports since I started my PhD <img class="emoji" title=":stuck_out_tongue_closed_eyes:" alt=":stuck_out_tongue_closed_eyes:" src="https://github.githubassets.com/images/icons/emoji/unicode/1f61d.png" height="20" width="20">).</p> 252 253<p style="color: red; font-weight: bold;"> 254â ï¸ My NUS staff email ([email protected]) is no longer active. 255Please contact me via my alumni email: [email protected]. 256</p> 257 258<center> 259<font size="6"> 260 261</font><center> 262<font size="6">
263<a href="mailto:%63%61%6F%79%75%68%6F%6E%67@%6E%75%73.%65%64%75.%73%67" title="email"><i class="fa-solid fa-envelope"></i></a> 264â 265<a href="https://scholar.google.com/citations?user=t7r7tjUAAAAJ" title="Google Scholar" rel="external nofollow noopener" target="_blank"><i class="ai ai-google-scholar"></i></a> 266â 267<a href="https://orcid.org/0000-0001-8099-0689" title="ORCID" rel="external nofollow noopener" target="_blank"><i class="ai ai-orcid"></i></a> 268â 269<a href="https://github.com/caoyuhong001#%20your%20GitHub%20user%20name" title="GitHub" rel="external nofollow noopener" target="_blank"><i class="fa-brands fa-github"></i></a> 270â 271<a href="https://www.linkedin.com/in/yuhong-cao" title="LinkedIn" rel="external nofollow noopener" target="_blank"><i class="fa-brands fa-linkedin"></i></a> 272</font> 273</center> 274 275<!-- 276Write your biography here. Tell the world about yourself. Link to your favorite [subreddit](http://reddit.com). You can put a picture in, too. The code is already in, just name your picture `prof_pic.jpg` and put it in the `img/` folder. 277 278Put your address / P.O. box / other info right below your picture. You can also disable any of these elements by editing `profile` property of the YAML header of your `_pages/about.md`. Edit `_bibliography/papers.bib` and Jekyll will render your [publications page](/al-folio/publications/) automatically. 279 280Link to your social media connections, too. This theme is set up to use [Font Awesome icons](https://fontawesome.com/) and [Academicons](https://jpswalsh.github.io/academicons/), like the ones below. Add your Facebook, Twitter, LinkedIn, Google Scholar, or just disable all of them. 281--> 282 283</center> 284</div> 285 286 <!-- News --> 287 288 <h2> 289 <a href="/news/" style="color: inherit">News</a> 290 </h2> 291 <div class="news"> 292 293 294 <div class="table-responsive" style="max-height: 20vw"> 295 <table class="table table-sm table-borderless"> 296 297 298 299 <tr> 300 <th scope="row" style="width: 20%">Jun 15, 2026</th> 301 <td> 302 303 Three of my co-author works on robust perceptive locomotion, efficient locomotion learning via action prior, and embodied navigation through imagination have been accepted to <a href="https://2026.ieee-iros.org/" rel="external nofollow noopener" target="_blank">IROS 2026</a>! 304 305 306 </td> 307 </tr> 308 309 <tr> 310 <th scope="row" style="width: 20%">May 06, 2026</th> 311 <td> 312 313 My co-author work which tackles multi-agent cooperative navigation by learning role play through option critic has been accepted by <a href="https://www.ieee-ras.org/publications/ra-l" rel="external nofollow noopener" target="_blank">RA-L</a>! 314 315 316 </td> 317 </tr> 318 319 <tr> 320 <th scope="row" style="width: 20%">May 02, 2026</th> 321 <td> 322 323 After six incredible years, today is my last day at NUS. This place shaped me into a PhD, a postdoc, and, more importantly, a better person. I am truly grateful to all the professors, mentors, colleagues, and friends who made this journey so meaningful! 324 325 326 </td> 327 </tr> 328 329 <tr> 330 <th scope="row" style="width: 20%">Jan 31, 2026</th> 331 <td> 332 333 My co-author work which improves our attention-based exploration pipeline through graph transformer and Kalman Filter has been accepted by <a href="https://2026.ieee-icra.org/" rel="external nofollow noopener" target="_blank">ICRA2026</a>! 334 335 336 </td> 337 </tr> 338 339 <tr> 340 <th scope="row" style="width: 20%">Dec 02, 2025</th> 341 <td> 342 343 I had the pleasure of giving an online talk on learning-based autonomous robot exploration, invited by Prof. <a href="https://mee.sustech.edu.cn/jiaozhiyuangong/3552.html" rel="external nofollow noopener" target="_blank">Boyu Zhou</a> from SUSTech. 344 345 346 </td> 347 </tr> 348 349 <tr> 350 <th scope="row" style="width: 20%">Oct 20, 2025</th> 351 <td> 352 353 Our work on scalable autonomous robot exploration based on expert guided reinforcement learning received the best paper award at the <a href="https://activep-ws.github.io/index.html" rel="external nofollow noopener" target="_blank">IROS2025 active perception workshop</a>! 354 355 356 </td> 357 </tr> 358 359 <tr> 360 <th scope="row" style="width: 20%">Oct 14, 2025</th> 361 <td> 362 363 I am honored to be a speaker at the <a href="https://macrai2025.github.io/" rel="external nofollow noopener" target="_blank">Multi-Agent Cooperative Systems and Swarm Robotics in the Era of Generative AI</a> workshop at <a href="https://www.iros25.org/" rel="external nofollow noopener" target="_blank">IROS2025</a>. 364 365 366 </td> 367 </tr> 368 369 <tr> 370 <th scope="row" style="width: 20%">Aug 02, 2025</th> 371 <td> 372 373 Three of my co-authored papers have been accepted to <a href="https://www.corl.org/" rel="external nofollow noopener" target="_blank">CORL2025</a>! Covering map prediction for learning-based robot exploration, test-time adaptation for vision-language-based outdoor animal search, and a novel approach to the general Traveling Salesman Problem. 374 375 376 </td> 377 </tr> 378 379 <tr> 380 <th scope="row" style="width: 20%">Apr 11, 2025</th> 381 <td> 382 383 My co-author work on bio-inspired torque control for locomotion is accepted by <a href="https://roboticsconference.org/" rel="external nofollow noopener" target="_blank">RSS</a>! 384 385 386 </td> 387 </tr> 388 389 <tr> 390 <th scope="row" style="width: 20%">Apr 01, 2025</th> 391 <td> 392 393 My co-author work on LLM-based multi-agent path-finding is accepted by <a href="https://2025.ijcnn.org/" rel="external nofollow noopener" target="_blank">IJCNN</a>! 394 395 396 </td> 397 </tr> 398 399 <tr> 400 <th scope="row" style="width: 20%">Jan 28, 2025</th> 401 <td> 402 403 Our paper on diffusion-based robot exploration is accepted by <a href="https://2025.ieee-icra.org/" rel="external nofollow noopener" target="_blank">ICRA2025</a>, along with two of my co-author works on multi-robot exploration with limited field-of-view and consensus for multi-robot path-finding inspired by sheaf theory! 404 405 406 </td> 407 </tr> 408 409 <tr> 410 <th scope="row" style="width: 20%">Jan 10, 2025</th> 411 <td> 412 413 My co-author work on learning-based heterogenous task allocation is accepted by <a href="https://www.ieee-ras.org/publications/ra-l" rel="external nofollow noopener" target="_blank">RA-L</a>! 414 415 416 </td> 417 </tr> 418 419 <tr> 420 <th scope="row" style="width: 20%">
420Nov 06, 2024</th> 421 <td> 422 423 My co-author work on hierarchical decision networks for robot exploration and navigation is accepted by <a href="https://www.ieee-ras.org/publications/ra-l" rel="external nofollow noopener" target="_blank">RA-L</a>! 424 425 426 </td> 427 </tr> 428 429 <tr> 430 <th scope="row" style="width: 20%">Sep 05, 2024</th> 431 <td> 432 433 My co-author work on multi-robot adversarial search is accepted by <a href="https://www.corl.org/" rel="external nofollow noopener" target="_blank">CORL2024</a>! 434 435 436 </td> 437 </tr> 438 439 <tr> 440 <th scope="row" style="width: 20%">Aug 08, 2024</th> 441 <td> 442 443 My co-author work on low-cost multi-UAV system development is accepted by <a href="https://2024.iccas.org/" rel="external nofollow noopener" target="_blank">ICCAS2024</a>! 444 445 446 </td> 447 </tr> 448 449 <tr> 450 <th scope="row" style="width: 20%">Jul 17, 2024</th> 451 <td> 452 453 My co-author work on multi-agent flocking is accepted as extended abstract by <a href="https://icra40.ieee.org/" rel="external nofollow noopener" target="_blank">ICRA@40</a>! 454 455 456 </td> 457 </tr> 458 459 <tr> 460 <th scope="row" style="width: 20%">Jul 02, 2024</th> 461 <td> 462 463 My co-author paper on multi-robot exploration under bandwidth constratint has been accepted by <a href="https://dars2024.engineering.cornell.edu/" rel="external nofollow noopener" target="_blank">DARS2024</a>! 464 465 466 </td> 467 </tr> 468 469 <tr> 470 <th scope="row" style="width: 20%">Jun 30, 2024</th> 471 <td> 472 473 My co-author paper on multi-robot exploration under limited communication range has been accepted by <a href="https://iros2024-abudhabi.org/" rel="external nofollow noopener" target="_blank">IROS2024</a>! 474 475 476 </td> 477 </tr> 478 479 <tr> 480 <th scope="row" style="width: 20%">Mar 08, 2024</th> 481 <td> 482 483 A new beginning! I am now a postdoc fellow at NUS ME. 484 485 486 </td> 487 </tr> 488 489 <tr> 490 <th scope="row" style="width: 20%">Feb 29, 2024</th> 491 <td> 492 493 Our paper on larg-scale robot exploration has been accepted by <a href="https://www.ieee-ras.org/publications/ra-l" rel="external nofollow noopener" target="_blank">RA-L</a>! 494 495 496 </td> 497 </tr> 498 499 <tr> 500 <th scope="row" style="width: 20%">Feb 28, 2024</th> 501 <td> 502 503 I obtained my PhD degree from NUS. Thanks to all who support me along this wonderful journey! 504 505 506 </td> 507 </tr> 508 509 <tr> 510 <th scope="row" style="width: 20%">Jan 15, 2024</th> 511 <td> 512 513 I passed the oral defense for my PhD thesis. A lot of thanks to my supervisor <a href="https://www.marmotlab.org/bio.html" rel="external nofollow noopener" target="_blank">Prof. Guillaume Sartoretti</a> and my examiners: <a href="https://cde.nus.edu.sg/me/staff/ang-jr-marcelo-h/" rel="external nofollow noopener" target="_blank">Prof. Marcelo Ang</a> and <a href="https://cde.nus.edu.sg/me/staff/chen-chao-yu-peter/" rel="external nofollow noopener" target="_blank">Prof. Chen Chao Yu Peter</a>! 514 515 516 </td> 517 </tr> 518 519 <tr> 520 <th scope="row" style="width: 20%">Sep 22, 2023</th> 521 <td> 522 523 My co-author paper on multi-agent informative path planning is accepted by <a href="https://sites.bu.edu/mrs2023/" rel="external nofollow noopener" target="_blank">MRS2023</a>! 524 525 526 </td> 527 </tr> 528 529 <tr> 530 <th scope="row" style="width: 20%">Aug 31, 2023</th> 531 <td> 532 533 My co-author paper on navigation in unknown environments is accepted by <a href="https://www.corl2023.org/" rel="external nofollow noopener" target="_blank">CoRL2023</a>! 534 535 536 </td> 537 </tr> 538 539 <tr> 540 <th scope="row" style="width: 20%">Jun 22, 2023</th> 541 <td> 542 543 My co-author paper on persistent monitoring is accepted by <a href="https://ieee-iros.org/" rel="external nofollow noopener" target="_blank">IROS2023</a>! 544 545 546 </td> 547 </tr> 548 549 <tr> 550 <th scope="row" style="width: 20%">Jan 17, 2023</th> 551 <td> 552 553 Our paper on autonomous robot exploration is accepted by <a href="https://www.icra2023.org/" rel="external nofollow noopener" target="_blank">ICRA2023</a>! 554 555 556 </td> 557 </tr> 558 559 <tr> 560 <th scope="row" style="width: 20%">Nov 21, 2022</th> 561 <td> 562 563 Our team wins a golden prize in <a href="https://cy.ncss.cn/en/" rel="external nofollow noopener" target="_blank">The 8th China International College Studentsâ âInternet+â Innovation and Entrepreneurship Competition</a>! Our project stands out as the final 200 golden-prize projects from more than one million teams. 564 565 566 </td> 567 </tr> 568 569 <tr> 570 <th scope="row" style="width: 20%">Sep 10, 2022</th> 571 <td> 572 573 Two papers accepted on the same day! My first paper on multiple traveling salesmen problem is accepted by <a href="https://dars2022.org/#/" rel="external nofollow noopener" target="_blank">DARS2022</a>. Another paper on informative path planning is accepted by <a href="https://corl2022.org/" rel="external nofollow noopener" target="_blank">CoRL2022</a>. 574 575 576 </td> 577 </tr> 578 579 <tr> 580 <th scope="row" style="width: 20%">Feb 28, 2022</th> 581 <td> 582 583 I transfered from master of engineering to dotor of philosophy! Sadly I can not be a âmasterâ anymore <img class="emoji" title=":smile:" alt=":smile:" src="https://github.githubassets.com/images/icons/emoji/unicode/1f604.png" height="20" width="20">. 584 585 586 </td> 587 </tr> 588 589 <tr> 590 <th scope="row" style="width: 20%">Aug 03, 2020</th> 591 <td> 592 593 I started my master of engineering in <a href="https://cde.nus.edu.sg/me/" rel="external nofollow noopener" target="_blank">Department of Mechanical Engineering, National University of Singapore</a>, under the supervision of <a href="https://www.marmotlab.org/" rel="external nofollow noopener" target="_blank">Dr. Guillaume Sartoretti</a>. 594 595 596 </td> 597 </tr> 598 599 <tr> 600 <th scope="row" style="width: 20%">Jun 30, 2020</th> 601 <td> 602 603 I received my bachelor degree in mechanical engineering from <a href="http://www.me.buaa.edu.cn/" rel="external nofollow noopener" target="_blank">
603School of Mechanical Engineering and Automation, Beihang University</a>. I am also awarded as the outstanding graduate of Beihang University! 604 605 606 </td> 607 </tr> 608 609 <tr> 610 <th scope="row" style="width: 20%">Sep 01, 2016</th> 611 <td> 612 613 I started my bachelor in mechanical engineering at <a href="http://www.me.buaa.edu.cn/" rel="external nofollow noopener" target="_blank">School of Mechanical Engineering and Automation, Beihang University</a>. 614 615 616 </td> 617 </tr> 618 619 </table> 620 </div> 621 622</div> 623 624 625 626 <!-- Latest posts --> 627 628 629 <!-- Selected papers --> 630 631 <h2> 632 <a href="/publications/" style="color: inherit">Selected Publications</a> 633 </h2> 634 <div class="publications"> 635 <ol class="bibliography"> 636<li> 637<div class="row"> 638 639 <div class="col col-sm-2 abbr"> 640 641 642 643 644 645<figure> 646 <picture> 647 <!-- Auto scaling with imagemagick --> 648 <!-- 649 See https://www.debugbear.com/blog/responsive-images#w-descriptors-and-the-sizes-attribute and 650 https://developer.mozilla.org/en-US/docs/Learn/HTML/Multimedia_and_embedding/Responsive_images for info on defining 'sizes' for responsive images 651 --> 652 653 <source class="responsive-img-srcset" srcset="/assets/img/publication_preview/header-480.webp 480w,/assets/img/publication_preview/header-800.webp 800w,/assets/img/publication_preview/header-1400.webp 1400w," sizes="200px" type="image/webp"> 654 655 <img src="/assets/img/publication_preview/header.gif" class="preview z-depth-1 rounded" width="100%" height="auto" alt="header.gif" data-zoomable loading="eager" onerror="this.onerror=null; $('.responsive-img-srcset').remove();"> 656 </source></picture> 657 658 659</figure> 660 661 662 663 </div> 664 665 666 <!-- Entry bib key --> 667 <div id="cao2025header" class="col-sm-8"> 668 <!-- Title --> 669 <div class="title">HEADER: Hierarchical Robot Exploration via Attention-Based Deep Reinforcement Learning with Expert-Guided Reward</div> 670 <!-- Author --> 671 <div class="author"> 672 673 674 675 <em>Yuhong 676 Cao</em>, Yizhuo 677 Wang, Jingsong 678 Liang, Shuhao 679 Liao, Yifeng 680 Zhang, Peizhuo 681 Li, and Guillaume 682 Sartoretti 683 </div> 684 685 <!-- Journal/Book title and date --> 686 687 688 689 690 691 692 693 694 695 696 <div class="periodical"> 697 <em>Under review</em>, 2025 698 </div> 699 <div class="periodical"> 700 701 </div> 702 703 <!-- Links/Buttons --> 704 <div class="links"> 705 706 707 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 708 709 710 711 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 712 713 714 715 716 <a href="https://arxiv.org/pdf/2510.15679" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">PDF</a> 717 718 719 720 721 <a href="https://drive.google.com/file/d/1aG3a5452QwMNHUftXHqKqVALj2yP-7KC/view?usp=sharing" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Video</a> 722 723 724 725 <a href="https://github.com/marmotlab/ARiADNE2-ROS-Planner" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> 726 727 728 729 730 </div> 731 732 733 734 735 736 737 738 739 740 741 <!-- Hidden abstract block --> 742 <div class="abstract hidden"> 743 <p>This work pushes the boundaries of learning-based methods in autonomous robot exploration in terms of environmental scale and exploration efficiency. We present HEADER, an attention-based reinforcement learning approach with hierarchical graphs for efficient exploration in large-scale environments. HEADER follows existing conventional methods to construct hierarchical representations for the robot belief/map, but further designs a novel community-based algorithm to construct and update a global graph, which remains fully incremental, shape-adaptive, and operates with linear complexity. Building upon attention-based networks, our planner finely reasons about the nearby belief within the local range while coarsely leveraging distant information at the global scale, enabling next-best-viewpoint decisions that consider multi-scale spatial dependencies. Beyond novel map representation, we introduce a parameter-free privileged reward that significantly improves model performance and produces near-optimal exploration behaviors, by avoiding training objective bi
743as caused by handcrafted reward shaping. In simulated challenging, large-scale exploration scenarios, HEADER demonstrates better scalability than most existing learning and non-learning methods, while achieving a significant improvement in exploration efficiency (up to 20%) over state-of-the-art baselines. We also deploy HEADER on hardware and validate it in complex, large-scale real-life scenarios, including a 300m*230m campus environment.</p> 744 </div> 745 746 747 748 <!-- Hidden bibtex block --> 749 <div class="bibtex hidden"> 750 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">cao2025header</span><span class="p">,</span> 751 <span class="na">title</span> <span class="p">=</span> <span class="s">{HEADER: Hierarchical Robot Exploration via Attention-Based Deep Reinforcement Learning with Expert-Guided Reward}</span><span class="p">,</span> 752 <span class="na">author</span> <span class="p">=</span> <span class="s">{Cao, Yuhong and Wang, Yizhuo and Liang, Jingsong and Liao, Shuhao and Zhang, Yifeng and Li, Peizhuo and Sartoretti, Guillaume}</span><span class="p">,</span> 753 <span class="na">journal</span> <span class="p">=</span> <span class="s">{Under review}</span><span class="p">,</span> 754 <span class="na">year</span> <span class="p">=</span> <span class="s">{2025}</span> 755<span class="p">}</span></code></pre></figure> 756 </div> 757 758 759 760 </div> 761</div> 762</li> 763<li> 764<div class="row"> 765 766 <div class="col col-sm-2 abbr"> 767 768 769 770 771 772<figure> 773 <picture> 774 <!-- Auto scaling with imagemagick --> 775 <!-- 776 See https://www.debugbear.com/blog/responsive-images#w-descriptors-and-the-sizes-attribute and 777 https://developer.mozilla.org/en-US/docs/Learn/HTML/Multimedia_and_embedding/Responsive_images for info on defining 'sizes' for responsive images 778 --> 779 780 <source class="responsive-img-srcset" srcset="/assets/img/publication_preview/dare-480.webp 480w,/assets/img/publication_preview/dare-800.webp 800w,/assets/img/publication_preview/dare-1400.webp 1400w," sizes="200px" type="image/webp"> 781 782 <img src="/assets/img/publication_preview/dare.gif" class="preview z-depth-1 rounded" width="100%" height="auto" alt="dare.gif" data-zoomable loading="eager" onerror="this.onerror=null; $('.responsive-img-srcset').remove();"> 783 </source></picture> 784 785 786</figure> 787 788 789 790 </div> 791 792 793 <!-- Entry bib key --> 794 <div id="cao2025dare" class="col-sm-8"> 795 <!-- Title --> 796 <div class="title">DARE: Diffusion Policy for Autonomous Robot Exploration</div> 797 <!-- Author --> 798 <div class="author"> 799 800 801 802 <em>Yuhong 803 Cao</em>, Jeric 804 Lew, Jingsong 805 Liang, Jin 806 Cheng, and Guillaume 807 Sartoretti 808 </div> 809 810 <!-- Journal/Book title and date --> 811 812 813 814 815 816 817 818 819 820 821 <div class="periodical"> 822 <em>2025 IEEE International Conference on Robotics and Automation (ICRA)</em>, 2025 823 </div> 824 <div class="periodical"> 825 826 </div> 827 828 <!-- Links/Buttons --> 829 <div class="links"> 830 831 832 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 833 834 835 836 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 837 838 839 840 841 <a href="https://arxiv.org/pdf/2410.16687" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">PDF</a> 842 843 844 845 846 <a href="https://drive.google.com/file/d/1fMzJoOQKK_nfPjQDta6mOIjewsQhN4rB/view?usp=drive_link" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Video</a> 847 848 849 850 <a href="https://github.com/marmotlab/DARE" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> 851 852 853 854 855 </div> 856 857 858 859 860 861 862 863 864 865 866 <!-- Hidden abstract block --> 867 <div class="abstract hidden"> 868 <p>
868Autonomous robot exploration requires a robot to efficiently explore and map unknown environments. Compared to conventional methods that can only optimize paths based on the current robot belief, learning-based methods show the potential to achieve improved performance by drawing on past experiences to reason about unknown areas. In this paper, we propose DARE, a novel generative approach that leverages diffusion models trained on expert demonstrations, which can explicitly generate an exploration path through one-time inference. We build DARE upon an attention-based encoder and a diffusion policy model, and introduce ground truth optimal demonstrations for training to learn better patterns for exploration. The trained planner can reason about the partial belief to recognize the potential structure in unknown areas and consider these areas during path planning. Our experiments demonstrate that DARE achieves on-par performance with both conventional and learning-based state-of-the-art exploration planners, as well as good generalizability in both simulations and real-life scenarios.</p> 869 </div> 870 871 872 873 <!-- Hidden bibtex block --> 874 <div class="bibtex hidden"> 875 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">cao2025dare</span><span class="p">,</span> 876 <span class="na">title</span> <span class="p">=</span> <span class="s">{DARE: Diffusion Policy for Autonomous Robot Exploration}</span><span class="p">,</span> 877 <span class="na">author</span> <span class="p">=</span> <span class="s">{Cao, Yuhong and Lew, Jeric and Liang, Jingsong and Cheng, Jin and Sartoretti, Guillaume}</span><span class="p">,</span> 878 <span class="na">journal</span> <span class="p">=</span> <span class="s">{2025 IEEE International Conference on Robotics and Automation (ICRA)}</span><span class="p">,</span> 879 <span class="na">year</span> <span class="p">=</span> <span class="s">{2025}</span><span class="p">,</span> 880 <span class="na">publisher</span> <span class="p">=</span> <span class="s">{IEEE}</span> 881<span class="p">}</span></code></pre></figure> 882 </div> 883 884 885 886 </div> 887</div> 888</li> 889<li> 890<div class="row"> 891 892 <div class="col col-sm-2 abbr"> 893 894 895 896 897 898<figure> 899 <picture> 900 <!-- Auto scaling with imagemagick --> 901 <!-- 902 See https://www.debugbear.com/blog/responsive-images#w-descriptors-and-the-sizes-attribute and 903 https://developer.mozilla.org/en-US/docs/Learn/HTML/Multimedia_and_embedding/Responsive_images for info on defining 'sizes' for responsive images 904 --> 905 906 <source class="responsive-img-srcset" srcset="/assets/img/publication_preview/drl_exploration-480.webp 480w,/assets/img/publication_preview/drl_exploration-800.webp 800w,/assets/img/publication_preview/drl_exploration-1400.webp 1400w," sizes="200px" type="image/webp"> 907 908 <img src="/assets/img/publication_preview/drl_exploration.gif" class="preview z-depth-1 rounded" width="100%" height="auto" alt="drl_exploration.gif" data-zoomable loading="eager" onerror="this.onerror=null; $('.responsive-img-srcset').remove();"> 909 </source></picture> 910 911 912</figure> 913 914 915 916 </div> 917 918 919 <!-- Entry bib key --> 920 <div id="cao2024deep" class="col-sm-8"> 921 <!-- Title --> 922 <div class="title">Deep Reinforcement Learning-based Large-scale Robot Exploration</div> 923 <!-- Author --> 924 <div class="author"> 925 926 927 928 <em>Yuhong 929 Cao</em>, Rui 930 Zhao, Yizhuo 931 Wang, Bairan 932 Xiang, and Guillaume 933 Sartoretti 934 </div> 935 936 <!-- Journal/Book title and date --> 937 938 939 940 941 942 943 944 945 946 947 <div class="periodical"> 948 <em>IEEE Robotics and Automation Letters</em>, 2024 949 </div> 950 <div class="periodical"> 951 952 </div> 953 954 <!-- Links/Buttons --> 955 <div class="links"> 956 957 958 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 959 960 961 962 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 963 964 965 966 <a href="https://ieeexplore.ieee.org/abstract/document/10476684" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">HTML</a> 967 968 969 970 971 <a href="https://arxiv.org/pdf/2403.10833" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">PDF</a> 972 973 974 975 976 <a href="https://drive.google.com/file/d/1nyVJnGOX2kv8Wsm2hLWIDGGKkhXAI4Tu/view?usp=sharing" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Video</a> 977 978 979 980 <a href="https://github.com/marmotlab/large-scale-DRL-exploration" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> 981 982 983 984 985 </div> 986 987 988 989 990 991 992 993 994 995 996 <!-- Hidden abstract block --> 997 <div class="abstract hidden"> 998 <p>In this work, we propose a deep reinforcement learning (DRL) based reactive planner to solve large-scale Lidar-based autonomous robot exploration problems in 2D action space. Our DRL-based planner allows the agent to reactively plan its exploration path by making implicit predictions about unknown areas, based on a learned estimation of the underlying transition model of the environment. To this end, our approach relies on learned attention mechanisms for their powerful ability to capture long-term dependencies at different spatial scales to reason about the robotâs entire belief over known areas. Our approach relies on ground truth information (i.e., privileged learning) to guide the environment estimation during training, as well as on a graph rarefaction algorithm, which allows models trained in small-scale environments to scale to large-scale ones. Simulation results show that our model exhibits better exploration efficie
998ncy (12% in path length, 6% in makespan) and lower planning time (60%) than the state-of-the-art planners in a 130mx100m benchmark scenario. We also validate our learned model on hardware.</p> 999 </div> 1000 1001 1002 1003 <!-- Hidden bibtex block --> 1004 <div class="bibtex hidden"> 1005 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@article</span><span class="p">{</span><span class="nl">cao2024deep</span><span class="p">,</span> 1006 <span class="na">title</span> <span class="p">=</span> <span class="s">{Deep Reinforcement Learning-based Large-scale Robot Exploration}</span><span class="p">,</span> 1007 <span class="na">author</span> <span class="p">=</span> <span class="s">{Cao, Yuhong and Zhao, Rui and Wang, Yizhuo and Xiang, Bairan and Sartoretti, Guillaume}</span><span class="p">,</span> 1008 <span class="na">journal</span> <span class="p">=</span> <span class="s">{IEEE Robotics and Automation Letters}</span><span class="p">,</span> 1009 <span class="na">year</span> <span class="p">=</span> <span class="s">{2024}</span><span class="p">,</span> 1010 <span class="na">publisher</span> <span class="p">=</span> <span class="s">{IEEE}</span> 1011<span class="p">}</span></code></pre></figure> 1012 </div> 1013 1014 1015 1016 </div> 1017</div> 1018</li> 1019<li> 1020<div class="row"> 1021 1022 <div class="col col-sm-2 abbr"> 1023 1024 1025 1026 1027 1028<figure> 1029 <picture> 1030 <!-- Auto scaling with imagemagick --> 1031 <!-- 1032 See https://www.debugbear.com/blog/responsive-images#w-descriptors-and-the-sizes-attribute and 1033 https://developer.mozilla.org/en-US/docs/Learn/HTML/Multimedia_and_embedding/Responsive_images for info on defining 'sizes' for responsive images 1034 --> 1035 1036 <source class="responsive-img-srcset" srcset="/assets/img/publication_preview/ariadne-480.webp 480w,/assets/img/publication_preview/ariadne-800.webp 800w,/assets/img/publication_preview/ariadne-1400.webp 1400w," sizes="200px" type="image/webp"> 1037 1038 <img src="/assets/img/publication_preview/ariadne.gif" class="preview z-depth-1 rounded" width="100%" height="auto" alt="ariadne.gif" data-zoomable loading="eager" onerror="this.onerror=null; $('.responsive-img-srcset').remove();"> 1039 </source></picture> 1040 1041 1042</figure> 1043 1044 1045 1046 </div> 1047 1048 1049 <!-- Entry bib key --> 1050 <div id="cao2023ariadne" class="col-sm-8"> 1051 <!-- Title --> 1052 <div class="title">Ariadne: A reinforcement learning approach using attention-based deep networks for exploration</div> 1053 <!-- Author --> 1054 <div class="author"> 1055 1056 1057 1058 <em>Yuhong 1059 Cao</em>, Tianxiang 1060 Hou, Yizhuo 1061 Wang, Xian 1062 Yi, and Guillaume 1063 Sartoretti 1064 </div> 1065 1066 <!-- Journal/Book title and date --> 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 <div class="periodical"> 1078 <em>In 2023 IEEE International Conference on Robotics and Automation (ICRA)</em> , 2023 1079 </div> 1080 <div class="periodical"> 1081 1082 </div> 1083 1084 <!-- Links/Buttons --> 1085 <div class="links"> 1086 1087 1088 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 1089 1090 1091 1092 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 1093 1094 1095 1096 <a href="https://ieeexplore.ieee.org/abstract/document/10160565" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">HTML</a> 1097 1098 1099 1100 1101 <a href="https://arxiv.org/pdf/2301.11575" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">PDF</a> 1102 1103 1104 1105 1106 <a href="https://drive.google.com/file/d/1h0p8d9Ry96FJGYI0mK1aNYicDCS7UPzs/view?usp=sharing" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Video</a> 1107 1108 1109 1110 <a href="https://github.com/marmotlab/ARiADNE" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> 1111 1112 1113 1114 1115 </div> 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 <!-- Hidden abstract block --> 1127 <div class="abstract hidden"> 1128 <p>In autonomous robot exploration tasks, a mobile robot needs to actively explore and map an unknown environment as fast as possible. Since the environment is being revealed during exploration, the robot needs to frequently re-plan its path online, as new information is acquired by onboard sensors and used to update its partial map. While state-of-the-art exploration planners are frontier- and sampling-based, encouraged by the recent development in deep reinforcement learning (DRL), we propose ARiADNE, an attention-based neural approach to obtain real-time, non-myopic path planning for autonomous exploration. ARiADNE is able to learn dependencies at multiple spatial scales between areas of the agentâs partial map, and implicitly predict potential gains associated with exploring those areas. This allows the agent to sequence movement actions that balance the natural trade-off between exploitation/refinement of the map in known areas and exploration of new areas. We experimentally demonstrate that our method outperforms both learning and non-learning state-of-the-art baselines in terms of average trajectory length to complete exploration in hundreds of simplified 2D indoor scenarios. We further validate our approach in high-fidelity Robot Operating System (ROS) simulations, where we consider a real sensor model and a realistic low-level motion controller, toward deployment on real robots.</p> 1129 </div> 1130 1131 1132 1133 <!-- Hidden bibtex block --> 1134 <div class="bibtex hidden"> 1135 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">cao2023ariadne</span><span class="p">,</span> 1136 <span class="na">title</span> <span class="p">=</span> <span class="s">
1136{Ariadne: A reinforcement learning approach using attention-based deep networks for exploration}</span><span class="p">,</span> 1137 <span class="na">author</span> <span class="p">=</span> <span class="s">{Cao, Yuhong and Hou, Tianxiang and Wang, Yizhuo and Yi, Xian and Sartoretti, Guillaume}</span><span class="p">,</span> 1138 <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{2023 IEEE International Conference on Robotics and Automation (ICRA)}</span><span class="p">,</span> 1139 <span class="na">pages</span> <span class="p">=</span> <span class="s">{10219--10225}</span><span class="p">,</span> 1140 <span class="na">year</span> <span class="p">=</span> <span class="s">{2023}</span><span class="p">,</span> 1141 <span class="na">organization</span> <span class="p">=</span> <span class="s">{IEEE}</span> 1142<span class="p">}</span></code></pre></figure> 1143 </div> 1144 1145 1146 1147 </div> 1148</div> 1149</li> 1150<li> 1151<div class="row"> 1152 1153 <div class="col col-sm-2 abbr"> 1154 1155 1156 1157 1158 1159<figure> 1160 <picture> 1161 <!-- Auto scaling with imagemagick --> 1162 <!-- 1163 See https://www.debugbear.com/blog/responsive-images#w-descriptors-and-the-sizes-attribute and 1164 https://developer.mozilla.org/en-US/docs/Learn/HTML/Multimedia_and_embedding/Responsive_images for info on defining 'sizes' for responsive images 1165 --> 1166 1167 <source class="responsive-img-srcset" srcset="/assets/img/publication_preview/catnipp-480.webp 480w,/assets/img/publication_preview/catnipp-800.webp 800w,/assets/img/publication_preview/catnipp-1400.webp 1400w," sizes="200px" type="image/webp"> 1168 1169 <img src="/assets/img/publication_preview/catnipp.gif" class="preview z-depth-1 rounded" width="100%" height="auto" alt="catnipp.gif" data-zoomable loading="eager" onerror="this.onerror=null; $('.responsive-img-srcset').remove();"> 1170 </source></picture> 1171 1172 1173</figure> 1174 1175 1176 1177 </div> 1178 1179 1180 <!-- Entry bib key --> 1181 <div id="cao2023catnipp" class="col-sm-8"> 1182 <!-- Title --> 1183 <div class="title">CAtNIPP: Context-aware attention-based network for informative path planning</div> 1184 <!-- Author --> 1185 <div class="author"> 1186 1187 1188 1189 <em>Yuhong 1190 Cao</em>, Yizhuo 1191 Wang, Apoorva 1192 Vashisth, Haolin 1193 Fan, and Guillaume Adrien 1194 Sartoretti 1195 </div> 1196 1197 <!-- Journal/Book title and date --> 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 <div class="periodical"> 1209 <em>In Conference on Robot Learning</em> , 2022 1210 </div> 1211 <div class="periodical"> 1212 1213 </div> 1214 1215 <!-- Links/Buttons --> 1216 <div class="links"> 1217 1218 1219 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 1220 1221 1222 1223 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 1224 1225 1226 1227 <a href="https://proceedings.mlr.press/v205/cao23b.html" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">HTML</a> 1228 1229 1230 1231 1232 <a href="https://proceedings.mlr.press/v205/cao23b/cao23b.pdf" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">PDF</a> 1233 1234 1235 1236 1237 <a href="https://drive.google.com/file/d/1yo5TrJCzEUE65jNeEgIOP_ON7uTVZtPv/view?usp=sharing" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Video</a> 1238 1239 1240 1241 <a href="https://github.com/marmotlab/CAtNIPP" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> 1242 1243 1244 1245 1246 </div> 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 <!-- Hidden abstract block --> 1258 <div class="abstract hidden"> 1259 <p>Informative path planning (IPP) is an NP-hard problem, which aims at planning a path allowing an agent to build an accurate belief about a quantity of interest throughout a given search domain, within constraints on resource budget (e.g., path length for robots with limited battery life). IPP requires frequent online replanning as this belief is updated with every new measurement (i.e., adaptive IPP), while balancing short-term exploitation and longer-term exploration to avoid suboptimal, myopic behaviors. Encouraged by the recent developments in deep reinforcement learning, we introduce CAtNIPP, a fully reactive, neural approach to the adaptive IPP problem. CAtNIPP relies on
1259self-attention for its powerful ability to capture 1260dependencies in data at multiple spatial scales. Specifically, our agent learns to form a context of its belief over the entire domain, which it uses to sequence local movement decisions 1261that optimize short- and longer-term search objectives. We experimentally demonstrate that CAtNIPP significantly outperforms state-of-the-art non-learning IPP solvers in terms of solution quality and computing time once trained, and present experimental results on hardware.</p> 1262 </div> 1263 1264 1265 1266 <!-- Hidden bibtex block --> 1267 <div class="bibtex hidden"> 1268 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">cao2023catnipp</span><span class="p">,</span> 1269 <span class="na">title</span> <span class="p">=</span> <span class="s">{CAtNIPP: Context-aware attention-based network for informative path planning}</span><span class="p">,</span> 1270 <span class="na">author</span> <span class="p">=</span> <span class="s">{Cao, Yuhong and Wang, Yizhuo and Vashisth, Apoorva and Fan, Haolin and Sartoretti, Guillaume Adrien}</span><span class="p">,</span> 1271 <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{Conference on Robot Learning}</span><span class="p">,</span> 1272 <span class="na">pages</span> <span class="p">=</span> <span class="s">{1928--1937}</span><span class="p">,</span> 1273 <span class="na">year</span> <span class="p">=</span> <span class="s">{2022}</span><span class="p">,</span> 1274 <span class="na">organization</span> <span class="p">=</span> <span class="s">{PMLR}</span> 1275<span class="p">}</span></code></pre></figure> 1276 </div> 1277 1278 1279 1280 </div> 1281</div> 1282</li> 1283<li> 1284<div class="row"> 1285 1286 <div class="col col-sm-2 abbr"> 1287 1288 1289 1290 1291 1292<figure> 1293 <picture> 1294 <!-- Auto scaling with imagemagick --> 1295 <!-- 1296 See https://www.debugbear.com/blog/responsive-images#w-descriptors-and-the-sizes-attribute and 1297 https://developer.mozilla.org/en-US/docs/Learn/HTML/Multimedia_and_embedding/Responsive_images for info on defining 'sizes' for responsive images 1298 --> 1299 1300 <source class="responsive-img-srcset" srcset="/assets/img/publication_preview/DAN-480.webp 480w,/assets/img/publication_preview/DAN-800.webp 800w,/assets/img/publication_preview/DAN-1400.webp 1400w," sizes="200px" type="image/webp"> 1301 1302 <img src="/assets/img/publication_preview/DAN.jpg" class="preview z-depth-1 rounded" width="100%" height="auto" alt="DAN.jpg" data-zoomable loading="eager" onerror="this.onerror=null; $('.responsive-img-srcset').remove();"> 1303 </source></picture> 1304 1305 1306</figure> 1307 1308 1309 1310 </div> 1311 1312 1313 <!-- Entry bib key --> 1314 <div id="cao2022dan" class="col-sm-8"> 1315 <!-- Title --> 1316 <div class="title">Dan: Decentralized attention-based neural network for the minmax multiple traveling salesman problem</div> 1317 <!-- Author --> 1318 <div class="author"> 1319 1320 1321 1322 <em>Yuhong 1323 Cao</em>, Zhanhong 1324 Sun, and Guillaume 1325 Sartoretti 1326 </div> 1327 1328 <!-- Journal/Book title and date --> 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 <div class="periodical"> 1340 <em>In International Symposium on Distributed Autonomous Robotic Systems</em> , 2022 1341 </div> 1342 <div class="periodical"> 1343 1344 </div> 1345 1346 <!-- Links/Buttons --> 1347 <div class="links"> 1348 1349 <a class="award btn btn-sm z-depth-0" role="button">Awareded</a> 1350 1351 1352 <a class="abstract btn btn-sm z-depth-0" role="button">Abs</a> 1353 1354 1355 1356 <a class="bibtex btn btn-sm z-depth-0" role="button">Bib</a> 1357 1358 1359 1360 <a href="https://link.springer.com/chapter/10.1007/978-3-031-51497-5_15" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">HTML</a> 1361 1362 1363 1364 1365 <a href="https://arxiv.org/pdf/2109.04205" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">PDF</a> 1366 1367 1368 1369 1370 <a href="https://drive.google.com/file/d/14oUQ-Ap5SYWLJPfWagcbmvhejsfU3JwA/view?usp=sharing" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Video</a> 1371 1372 1373 1374 <a href="https://github.com/marmotlab/DAN" class="btn btn-sm z-depth-0" role="button" rel="external nofollow noopener" target="_blank">Code</a> 1375 1376 1377 1378 1379 </div> 1380 1381 1382 1383 1384 1385 1386 1387 1388 <!-- Hidden Award block --> 1389 <div class="award hidden d-print-inline"> 1390 <p></p> 1391<p>Best student paper</p> 1392 1393 </div> 1394 1395 1396 1397 <!-- Hidden abstract block --> 1398 <div class="abstract hidden"> 1399 <p>The multiple traveling salesman problem (mTSP) is a well-known NP-hard problem with numerous real-world applications. In particular, this work addresses MinMax mTSP, where the objective is to minimize the max tour length among all agents. Many robotic deployments require recomputing potentially large mTSP instances frequently, making the natural trade-off between computing time and solution quality of great importance. However, exact and heuristic algorithms become inefficient as the number of cities increases, due to their computational complexity. Encouraged by the recent developments in deep reinforcement learning (dRL), this work approaches the mTSP as a cooperative task and introduces DAN, a decentralized attention-based neural method that aims at tackling this key trade-off. In DAN, agents learn fully decentralized policies to collaboratively construct a tour, by predicting each otherâs future decisions. Our model relies on attention mechanisms and is trained using multi-agent RL with parameter sharing, providing natural scalability to the numbers of agents and cities. Our experimental results on
1399small- to large-scale mTSP instances (50 to 1000 cities and 5 to 20 agents) show that DAN is able to match or outperform state-of-the-art solvers while keeping planning times low. In particular, given the same computation time budget, DAN outperforms all conventional and dRL-based baselines on larger-scale instances (more than 100 cities, more than 5 agents), and exhibits enhanced agent collaboration.</p> 1400 </div> 1401 1402 1403 1404 <!-- Hidden bibtex block --> 1405 <div class="bibtex hidden"> 1406 <figure class="highlight"><pre><code class="language-bibtex" data-lang="bibtex"><span class="nc">@inproceedings</span><span class="p">{</span><span class="nl">cao2022dan</span><span class="p">,</span> 1407 <span class="na">title</span> <span class="p">=</span> <span class="s">{Dan: Decentralized attention-based neural network for the minmax multiple traveling salesman problem}</span><span class="p">,</span> 1408 <span class="na">author</span> <span class="p">=</span> <span class="s">{Cao, Yuhong and Sun, Zhanhong and Sartoretti, Guillaume}</span><span class="p">,</span> 1409 <span class="na">booktitle</span> <span class="p">=</span> <span class="s">{International Symposium on Distributed Autonomous Robotic Systems}</span><span class="p">,</span> 1410 <span class="na">pages</span> <span class="p">=</span> <span class="s">{202--215}</span><span class="p">,</span> 1411 <span class="na">year</span> <span class="p">=</span> <span class="s">{2022}</span><span class="p">,</span> 1412 <span class="na">organization</span> <span class="p">=</span> <span class="s">{Springer}</span> 1413<span class="p">}</span></code></pre></figure> 1414 </div> 1415 1416 1417 1418 </div> 1419</div> 1420</li> 1421</ol> 1422</div> 1423 1424 1425 1426 <!-- Social --> 1427 1428 1429 <div class="picture"> 1430 1431 1432<figure> 1433 <picture> 1434 <!-- Auto scaling with imagemagick --> 1435 <!-- 1436 See https://www.debugbear.com/blog/responsive-images#w-descriptors-and-the-sizes-attribute and 1437 https://developer.mozilla.org/en-US/docs/Learn/HTML/Multimedia_and_embedding/Responsive_images for info on defining 'sizes' for responsive images 1438 --> 1439 1440 <source class="responsive-img-srcset" srcset="/assets/img/pic-480.webp 480w,/assets/img/pic-800.webp 800w,/assets/img/pic-1400.webp 1400w," sizes="95vw" type="image/webp"> 1441 1442 <img src="/assets/img/pic.jpg?fd5f6854f51f5bfdda639391077aad8c" width="100%" height="auto" loading="eager" onerror="this.onerror=null; $('.responsive-img-srcset').remove();"> 1443 </source></picture> 1444 1445 1446</figure> 1447 1448 <h6> 1449 <a style="align-items: center; display: flex; justify-content: center;">This is my hometown Chongqing. It is situated at the confluence of the Yangtze River and the Jialing River.</a> 1450 </h6> 1451 </div> 1452 </article> 1453</div> 1454 1455 1456 </div> 1457 1458 <!-- Footer --> 1459 1460 <footer class="fixed-bottom" role="contentinfo"> 1461 <div class="container mt-0"> 1462 © Copyright 2026 1463 Yuhong 1464 1465 Cao. Powered by <a href="https://jekyllrb.com/" target="_blank" rel="external nofollow noopener">Jekyll</a> with <a href="https://github.com/alshedivat/al-folio" rel="external nofollow noopener" target="_blank">al-folio</a> theme. 1466 1467 1468 1469 Last updated: June 20, 2026. 1470 1471 </div> 1472 </footer> 1473 1474 1475 1476 <!-- JavaScripts --> 1477 <!-- jQuery -->
1478<script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/jquery.min.js" integrity="sha256-/xUj+3OJU5yExlq6GSYGSHk7tPXikynS7ogEvDej/m4=" crossorigin="anonymous"></script>
1478 1479 1480 <!-- Bootsrap & MDB scripts -->
1481<script src="/assets/js/bootstrap.bundle.min.js"></script>
1481 1482<!--
1482<script src="/assets/js/mdb.min.js"></script>
1482 -->
1483<script src="https://cdn.jsdelivr.net/npm/[email protected]/js/mdb.min.js" integrity="sha256-NdbiivsvWt7VYCt6hYNT3h/th9vSTL4EDWeGs5SN3DA=" crossorigin="anonymous"></script>
1483 1484 1485 1486 <!-- Masonry & imagesLoaded --> 1487
1487<script defer src="https://cdn.jsdelivr.net/npm/[email protected]/dist/masonry.pkgd.min.js" integrity="sha256-Nn1q/fx0H7SNLZMQ5Hw5JLaTRZp0yILA/FRexe19VdI=" crossorigin="anonymous"></script>
1487 1488
1488<script defer src="https://cdn.jsdelivr.net/npm/[email protected]/imagesloaded.pkgd.min.js" integrity="sha256-htrLFfZJ6v5udOG+3kNLINIKh2gvoKqwEhHYfTTMICc=" crossorigin="anonymous"></script>
1488 1489
1489<script defer src="/assets/js/masonry.js" type="text/javascript"></script>
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 <!-- Medium Zoom JS --> 1511
1511<script defer src="https://cdn.jsdelivr.net/npm/[email protected]/dist/medium-zoom.min.js" integrity="sha256-ZgMyDAIYDYGxbcpJcfUnYwNevG/xi9OHKaR/8GK+jWc=" crossorigin="anonymous"></script>
1511 1512
1512<script defer src="/assets/js/zoom.js?85ddb88934d28b74e78031fd54cf8308"></script>
1512 1513 1514 1515 1516<!-- Bootstrap Table --> 1517 1518 1519<!-- Load Common JS -->
1520<script src="/assets/js/no_defer.js?2781658a0a2b13ed609542042a859126"></script>
vendor: 1 bytes, line 1520
1520
1521<script defer src="/assets/js/common.js?06cae41083477f121be8cd9797ad8e2f"></script>
vendor: 1 bytes, line 1521
1521
1522<script defer src="/assets/js/copy_code.js?12775fdf7f95e901d7119054556e495f" type="text/javascript"></script>
1522 1523 1524<!-- Jupyter Open External Links New Tab -->
1525<script defer src="/assets/js/jupyter_new_tab.js?d9f17b6adc2311cbabd747f4538bb15f"></script>
1525 1526 1527 1528 1529 1530
1530<script async src="https://d1bxh8uas1mnw7.cloudfront.net/assets/embed.js"></script>
1530 1531 1532 1533
1533<script async src="https://badge.dimensions.ai/badge.js"></script>
1533 1534 1535 1536 1537 1538 <!-- MathJax --> 1539
1539<script type="text/javascript"> 1540 window.MathJax = { 1541 tex: { 1542 tags: 'ams', 1543 }, 1544 }; 1545 </script>
1545 1546
1546<script defer type="text/javascript" id="MathJax-script" src="https://cdn.jsdelivr.net/npm/[email protected]/es5/tex-mml-chtml.min.js" integrity="sha256-rjmgmaB99riUNcdlrDtcAiwtLIojSxNyUFdl+Qh+rB4=" crossorigin="anonymous"></script>
1546 1547
1547<script defer src="https://cdnjs.cloudflare.com/polyfill/v3/polyfill.min.js?features=es6" crossorigin="anonymous"></script>
1547 1548 1549 1550 1551 1552 1553 1554 <!-- Global site tag (gtag.js) - Google Analytics -->
vendor: 66 bytes, lines 1554-1555
1554 1555 <script async src="https://www.googletagmanager.com/gtag/js?id=
1555G-2F7M79CZKB
vendor: 14 bytes, lines 1555-1556
1555"></script> 1556
1556<script> 1557 window.dataLayer = window.dataLayer || []; 1558 function gtag() { 1559 window.dataLayer.push(arguments); 1560 } 1561 gtag('js', new Date()); 1562 gtag('config', 'G-2F7M79CZKB'); 1563 </script>
1563 1564 1565 1566 1567 1568 1569 <!-- Scrolling Progress Bar --> 1570
1570<script type="text/javascript"> 1571 /* 1572 * This JavaScript code has been adapted from the article 1573 * https://css-tricks.com/reading-position-indicator/ authored by Pankaj Parashar, 1574 * published on the website https://css-tricks.com on the 7th of May, 2014. 1575 * Couple of changes were made to the original code to make it compatible 1576 * with the `al-foio` theme. 1577 */ 1578 const progressBar = $('#progress'); 1579 /* 1580 * We set up the bar after all elements are done loading. 1581 * In some cases, if the images in the page are larger than the intended 1582 * size they'll have on the page, they'll be resized via CSS to accomodate 1583 * the desired size. This mistake, however, breaks the computations as the 1584 * scroll size is computed as soon as the elements finish loading. 1585 * To account for this, a minimal delay was introduced before computing the 1586 * values. 1587 */ 1588 window.onload = function () { 1589 setTimeout(progressBarSetup, 50); 1590 }; 1591 /* 1592 * We set up the bar according to the browser. 1593 * If the browser supports the progress element we use that. 1594 * Otherwise, we resize the bar thru CSS styling 1595 */ 1596 function progressBarSetup() { 1597 if ('max' in document.createElement('progress')) { 1598 initializeProgressElement(); 1599 $(document).on('scroll', function () { 1600 progressBar.attr({ value: getCurrentScrollPosition() }); 1601 }); 1602 $(window).on('resize', initializeProgressElement); 1603 } else { 1604 resizeProgressBar(); 1605 $(document).on('scroll', resizeProgressBar); 1606 $(window).on('resize', resizeProgressBar); 1607 } 1608 } 1609 /* 1610 * The vertical scroll position is the same as the number of pixels that 1611 * are hidden from view above the scrollable area. Thus, a value > 0 is 1612 * how much the user has scrolled from the top 1613 */ 1614 function getCurrentScrollPosition() { 1615 return $(window).scrollTop(); 1616 } 1617 1618 function initializeProgressElement() { 1619 let navbarHeight = $('#navbar').outerHeight(true); 1620 $('body').css({ 'padding-top': navbarHeight }); 1621 $('progress-container').css({ 'padding-top': navbarHeight }); 1622 progressBar.css({ top: navbarHeight }); 1623 progressBar.attr({ 1624 max: getDistanceToScroll(), 1625 value: getCurrentScrollPosition(), 1626 }); 1627 } 1628 /* 1629 * The offset between the html document height and the browser viewport 1630 * height will be greater than zero if vertical scroll is possible. 1631 * This is the distance the user can scroll 1632 */ 1633 function getDistanceToScroll() { 1634 return $(document).height() - $(window).height(); 1635 } 1636 1637 function resizeProgressBar() { 1638 progressBar.css({ width: getWidthPercentage() + '%' }); 1639 } 1640 // The scroll ratio equals the percentage to resize the bar 1641 function getWidthPercentage() { 1642 return (getCurrentScrollPosition() / getDistanceToScroll()) * 100; 1643 } 1644 </script>
1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654
1654<script src="/assets/js/vanilla-back-to-top.min.js?f40d453793ff4f64e238e420181a1d17"></script>
1654 1655
1655<script> 1656 addBackToTop(); 1657 </script>
1657 1658 1659 1660 1661 1662
1662<script src="/assets/js/shortcut-key.js"></script>
1662 1663 </body> 1664</html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.