1<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" 2 "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> 3<!-- saved from url=(0069)http://www.cs.cmu.edu/~dchaplot/projects/neural-topological-slam.html --> 4 5<html xmlns="http://www.w3.org/1999/xhtml"> 6<!-- ======================================================================= --> 7 8<head> 9 <meta http-equiv="Content-Type" content="text/html; charset=us-ascii" /> 10
10<script type="text/javascript" id="www-widgetapi-script" src="./website_files/www-widgetapi.js" async=""> 11</script>
11 12
12<script src="./website_files/jsapi" type="text/javascript"> 13</script>
13 14
14<script type="text/javascript"> 15//<![CDATA[ 16 google.load("jquery", "1.3.2"); 17 //]]> 18 </script>
18 19
19<script type="text/javascript" charset="UTF-8" src="./website_files/jquery.min.js"> 20</script>
20 21 <style type="text/css"> 22/*<![CDATA[*/ 23 body { 24 font-family: "Titillium Web","HelveticaNeue-Light", "Helvetica Neue Light", "Helvetica Neue", Helvetica, Arial, "Lucida Grande", sans-serif; 25 font-weight:300; 26 font-size:18px; 27 margin-left: auto; 28 margin-right: auto; 29 width: 1100px; 30 } 31 32 h1 { 33 font-weight:300; 34 } 35 36 .disclaimerbox { 37 background-color: #eee; 38 border: 1px solid #eeeeee; 39 border-radius: 10px ; 40 -moz-border-radius: 10px ; 41 -webkit-border-radius: 10px ; 42 padding: 20px; 43 } 44 45 video.header-vid { 46 height: 140px; 47 border: 1px solid black; 48 border-radius: 10px ; 49 -moz-border-radius: 10px ; 50 -webkit-border-radius: 10px ; 51 } 52 53 img.header-img { 54 height: 140px; 55 border: 1px solid black; 56 border-radius: 10px ; 57 -moz-border-radius: 10px ; 58 -webkit-border-radius: 10px ; 59 } 60 61 img.rounded { 62 border: 1px solid #eeeeee; 63 border-radius: 10px ; 64 -moz-border-radius: 10px ; 65 -webkit-border-radius: 10px ; 66 } 67 68 a:link,a:visited 69 { 70 color: #1367a7; 71 text-decoration: none; 72 } 73 a:hover { 74 color: #208799; 75 } 76 77 td.dl-link { 78 height: 160px; 79 text-align: center; 80 font-size: 22px; 81 } 82 83 .layered-paper-big { /* modified from: http://css-tricks.com/snippets/css/layered-paper/ */ 84 box-shadow: 85 0px 0px 1px 1px rgba(0,0,0,0.35), /* The top layer shadow */ 86 5px 5px 0 0px #fff, /* The second layer */ 87 5px 5px 1px 1px rgba(0,0,0,0.35), /* The second layer shadow */ 88 10px 10px 0 0px #fff, /* The third layer */ 89 10px 10px 1px 1px rgba(0,0,0,0.35), /* The third layer shadow */ 90 15px 15px 0 0px #fff, /* The fourth layer */ 91 15px 15px 1px 1px rgba(0,0,0,0.35), /* The fourth layer shadow */ 92 20px 20px 0 0px #fff, /* The fifth layer */ 93 20px 20px 1px 1px rgba(0,0,0,0.35), /* The fifth layer shadow */ 94 25px 25px 0 0px #fff, /* The fifth layer */ 95 25px 25px 1px 1px rgba(0,0,0,0.35); /* The fifth layer shadow */ 96 margin-left: 10px; 97 margin-right: 45px; 98 } 99 100 101 .layered-paper { /* modified from: http://css-tricks.com/snippets/css/layered-paper/ */ 102 box-shadow: 103 0px 0px 1px 1px rgba(0,0,0,0.35), /* The top layer shadow */ 104 5px 5px 0 0px #fff, /* The second layer */ 105 5px 5px 1px 1px rgba(0,0,0,0.35), /* The second layer shadow */ 106 10px 10px 0 0px #fff, /* The third layer */ 107 10px 10px 1px 1px rgba(0,0,0,0.35); /* The third layer shadow */ 108 margin-top: 5px; 109 margin-left: 10px; 110 margin-right: 30px; 111 margin-bottom: 5px; 112 } 113 114 .vert-cent { 115 position: relative; 116 top: 50%; 117 transform: translateY(-50%); 118 } 119 120 hr 121 { 122 border: 0; 123 height: 1px; 124 background-image: linear-gradient(to right, rgba(0, 0, 0, 0), rgba(0, 0, 0, 0.75), rgba(0, 0, 0, 0)); 125 } 126 127 #authors td { 128 padding-bottom:5px; 129 padding-top:30px; 130 } 131 /*]]>*/ 132 </style><!-- ======================================================================= --> 133 <!-- Global site tag (gtag.js) - Google Analytics --> 134 135
135<script async="" src="./website_files/js" type="text/javascript"> 136</script>
136 137
137<script type="text/javascript"> 138//<![CDATA[ 139
vendor: 134 bytes, lines 139-143
139window.dataLayer = window.dataLayer || []; 140 function gtag(){dataLayer.push(arguments);} 141 gtag('js', new Date()); 142 143 gtag('config', '
143UA-114291442-6
vendor: 6 bytes, lines 143-144
143'); 144
144//]]> 145 </script>
145 146
146<script type="text/javascript" src="./website_files/hidebib.js"> 147</script>
147 148 <link href="./website_files/css" rel="stylesheet" type="text/css" /> 149 <meta http-equiv="X-UA-Compatible" content="IE=edge" /> 150 <link rel="icon" type="image/png" href="" /> 151 152 <title>Estimating Perceptual Uncertainty to Predict Robust Motion Plans</title> 153 <meta name="HandheldFriendly" content="True" /> 154 <meta name="viewport" content="width=device-width, initial-scale=1.0" /> 155
155<script src="./website_files/iframe_api" type="text/javascript"> 156</script>
156 157</head> 158 159<body> 160 <br /> 161 162 163 <center> 164 <span style="font-size:44px;font-weight:bold;">Precise Mobile Manipulation<br/>of Small Everyday Objects</span> 165 </center> 166 <br /> 167 168 169 <table align="center" width="800px"> 170 <tbody> 171 <tr> 172 <td align="center" width="230px"> 173 <center> 174 <span style="font-size:22px">Arjun Gupta</span> 175 </center> 176 </td> 177 178 <td align="center" width="230px"> 179 <center> 180 <span style="font-size:22px">Rishik Sathua</span> 181 </center> 182 </td> 183 184 <td align="center" width="230px"> 185 <center> 186 <span style="font-size:22px">Saurabh Gupta</span> 187 </center> 188 </td> 189 </tr> 190 191 192 <tr> 193 <td align="center" width="230"> 194 <center> 195 <span style="font-size:20px">UIUC</span> 196 </center> 197 </td> 198 199 <td align="center" width="230"> 200 <center> 201 <span style="font-size:20px">UIUC</span> 202 </center> 203 </td> 204 205 <td align="center" width="230px"> 206 <center> 207 <span style="font-size:20px">UIUC</span> 208 </center> 209 </td> 210 </tr> 211 </tbody> 212 </table> 213 214 <center> 215 <div style="margin-top: 10px; margin-bottom: 10px;"> 216 <span style="font-size:25px; font-weight:bold;">RA-L 2026</span> 217 </div> 218 </center> 219 220 <table align="center" width="700px" style="padding-top: 15px"> 221 <tbody><tr> 222 <td align="center" width="200px"><center><span style="font-size:24px"><a href="https://arxiv.org/abs/2502.13964">Paper</a></span></center></td> 223 </tr><tr> 224 </tr></tbody></table> 225 226 <table align="center" width="300px" style="padding-top: 25px"> 227 <tbody> 228 <tr> 229 <td align="center" width="300px"><iframe width="700" height="400" src="https://www.youtube.com/embed/1BFclCB91G0" frameborder="0" allow="accelerometer; autoplay; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe></iframe> 230 </td> 231 </tr> 232 </tbody> 233 </table> 234 <br /> 235 236 <br /> 237 238 239 <div style="width:800px; margin:0 auto; text-align:justify"> 240 Many everyday mobile manipulation tasks require precise interaction with small objects, such as grasping a knob to open a cabinet or pressing a light switch. In this paper, we develop Servoing with Vision Models (SVM), a closed-loop framework that enables a mobile manipulator to tackle such precise tasks involving the manipulation of small objects. SVM uses state-of-the-art vision foundation models to generate 3D targets for visual servoing to enable diverse tasks in novel environments. Naively doing so fails because of occlusion by the end-effector. SVM mitigates this using vision models that out-paint the end-effector, thereby significantly enhancing target localization. We demonstrate that aided by out-painting methods, open-vocabulary object detectors can serve as a drop-in module for SVM to seek semantic targets (e.g. knobs) and point tracking methods can help SVM reliably pursue interaction sites indicated by user clicks. We conduct a large-scale evaluation spanning experiments in 10 novel environments across 6 buildings including 72 different object instances. SVM obtains a 71% zero-shot success rate on manipulating unseen objects in novel environments in the real world, outperforming an open-loop control method by an absolute 42% and an imitation learning baseline trained on 1000+ demonstrations also by an absolute success rate of 50%. 241 242 </div> 243 244 <div style="width:1000px; margin:0 auto; text-align:center"> 245 246 </div> 247 248 <br /> 249 250 <hr /> 251 252
252<script xml:space="preserve" language="JavaScript" type="text/javascript"> 253 <!--hideallbibs();--> 254 </script>
254 255</body> 256</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.