Skip to content

Commit dc2a0b2

Browse files
siminliu99siminliu99
committed
cv updated, talks and teaching consistent with CV, done for now
1 parent 7f26218 commit dc2a0b2

16 files changed

Lines changed: 51 additions & 18 deletions

‎_pages/about.md‎

Lines changed: 45 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -43,17 +43,39 @@ You can reach me at simin.liu.1314 -at- gmail dot com
4343

4444
<div class="portfolio-list">
4545

46-
<div class="portfolio-item">
47-
<h3>Higher-Quality Planning for Contact-Rich Manipulation</h3>
48-
<p> We built a planner that enables a bimanual system to move large, heavy objects using whole-arm contact. Unlike prior sampling-based approaches, which could produce whole-arm plans but at poor quality, this planner globally optimizes over grasp sequencing and in-grasp motion jointly. This joint optimization produces consistent, efficient plans suitable for hardware deployment and reinforcement learning. </p>
49-
<div class="portfolio-media">
50-
<!-- Replace src with your image or embed a video here -->
51-
<!-- <img src="/images/portfolio/contact_planning.png" alt="Contact-rich manipulation planning" /> -->
52-
<video autoplay loop muted playsinline preload="auto">
53-
<source src="/images/portfolio/crm.mp4" type="video/mp4">
54-
</video>
55-
<!-- <p class="portfolio-caption">Our method generates shorter, more direct plans than a state-of-the-art sampling-based planner.</p> -->
56-
<p class="portfolio-caption">Our method generates short, direct plans that leverage all manipulator surfaces, not just end-effectors.</p>
46+
<div class="portfolio-group">
47+
<h3 class="portfolio-group-title">Dexterous, Contact-Rich Manipulation</h3>
48+
<p class="portfolio-group-intro">Building learning and planning algorithms for dexterous, contact-rich manipulation, where the full arm is used to move objects, not just the end-effector. Contact-rich manipulation is more challenging and more expressive than pick-and-place.</p>
49+
50+
<div class="portfolio-subitem">
51+
<h4>Higher-Quality Model-Based Planning</h4>
52+
<p>We built a planner that enables a bimanual system to move large, heavy objects using whole-arm contact. Unlike prior sampling-based approaches, which could produce whole-arm plans but at poor quality, this planner globally optimizes over grasp sequencing and in-grasp motion jointly. This joint optimization produces consistent, efficient plans suitable for hardware deployment and reinforcement learning. </p>
53+
<div class="portfolio-media">
54+
<!-- <img src="/images/portfolio/safe_control_highdim.png" alt="Safe control for high-dimensional systems" /> -->
55+
<video autoplay loop muted playsinline preload="auto">
56+
<source src="/images/portfolio/crm.mp4" type="video/mp4">
57+
</video>
58+
<p class="portfolio-caption">Our method generates short, direct plans that leverage all manipulator surfaces, not just end-effectors.</p>
59+
</div>
60+
</div>
61+
62+
<div class="portfolio-subitem">
63+
<h4>Learning from Planner-Generated Demonstrations </h4>
64+
<p> (Ongoing work): Synthetic data avoids the cross-embodiment transfer issues of human data, and is therefore a promising additional data source for today's VLAs and RL algorithms. Teleoperation is also often awkward for contact-rich manipulation. Building on our planner for contact-rich manipulation, we're using its outputs as synthetic demonstrations for RL, and measuring how much they accelerate training and where the gains are largest. </p>
65+
<div class="portfolio-media">
66+
<div class="portfolio-grid-3x3">
67+
<img src="/images/grs_query_7.gif" alt="Planner demo" />
68+
<img src="/images/grs_query_9.gif" alt="Planner demo" />
69+
<img src="/images/grs_query_63.gif" alt="Planner demo" />
70+
<img src="/images/grs_query_66.gif" alt="Planner demo" />
71+
<img src="/images/grs_query_75.gif" alt="Planner demo" />
72+
<img src="/images/grs_query_78.gif" alt="Planner demo" />
73+
<img src="/images/grs_query_115.gif" alt="Planner demo" />
74+
<img src="/images/grs_query_161.gif" alt="Planner demo" />
75+
<img src="/images/grs_query_190.gif" alt="Planner demo" />
76+
</div>
77+
<p class="portfolio-caption">A sampling of planner-generated demonstrations for different (start, goal) queries.</p>
78+
</div>
5779
</div>
5880
</div>
5981

@@ -89,7 +111,7 @@ We synthesized robust-adaptive safety filters for nonlinear systems with unknown
89111

90112
<div class="portfolio-item">
91113
<h3>Multitask RL for Adaptive Locomotion</h3>
92-
<p>WModel-based methods and standard RL both struggle to generalize locomotion controllers to previously unseen disturbances. We develop a multitask model-based RL algorithm that trains an adaptable dynamics model on a few hours of domain-randomized data — scenarios like leg loss, terrain variation, and payload changes. We demonstrate a 3–8x increase in path-following reward over a no-adaptation baseline on unseen disturbances.</p>
114+
<p>Model-based methods and standard RL both struggle to generalize locomotion controllers to previously unseen disturbances. We develop a multitask model-based RL algorithm that trains an adaptable dynamics model on a few hours of domain-randomized data — scenarios like leg loss, terrain variation, and payload changes. We demonstrate a 3–8x increase in path-following reward over a no-adaptation baseline on unseen disturbances.</p>
93115
<div class="portfolio-media">
94116
<!-- <img src="/images/portfolio/locomotion.png" alt="Locomotion under disturbances" /> -->
95117
<video autoplay loop muted playsinline preload="auto">
@@ -126,6 +148,17 @@ We synthesized robust-adaptive safety filters for nonlinear systems with unknown
126148
width: 100%;
127149
display: block;
128150
}
151+
.portfolio-grid-3x3 {
152+
display: grid;
153+
grid-template-columns: repeat(3, 1fr);
154+
gap: 4px;
155+
}
156+
.portfolio-grid-3x3 img {
157+
width: 100%;
158+
height: auto;
159+
display: block;
160+
object-fit: cover;
161+
}
129162
.portfolio-caption {
130163
margin: 0;
131164
padding: 0.6rem 0.9rem;

‎_pages/cv.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -9,4 +9,4 @@ redirect_from:
99

1010
{% include base_path %}
1111

12-
<iframe src="/files/Simin_Liu_CV_03_02_26.pdf" width="100%" height="900px" style="border: none;"></iframe>
12+
<iframe src="/files/Simin_Liu_CV_04_24_26.pdf" width="100%" height="900px" style="border: none;"></iframe>

‎_talks/2026-02-01-manipulation-seminar.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
11
---
2-
title: "High-Performance Planning for Contact-Rich Manipulation"
2+
title: "Towards Higher-Quality Planning for Contact-Rich Manipulation"
33
collection: talks
44
type: "Talk"
55
permalink: /talks/2026-manipulation-seminar

‎_talks/2026-03-23-duke-robotics-seminar.md‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,10 +1,10 @@
11
---
2-
title: "High-Performance Planning for Contact-Rich Manipulation"
2+
title: "Towards Higher-Quality Planning for Contact-Rich Manipulation"
33
collection: talks
44
type: "Talk"
5-
permalink: /talks/2026-duke-robotics-seminar
6-
venue: "Robotics Seminar, Duke University"
5+
permalink: /talks/2026-duke-manipulation-seminar
6+
venue: "Manipulation Seminar, Duke University"
77
date: 2026-03-23
88
location: "Pittsburgh, PA"
99
---
10-
Invited talk at Duke Robotics Seminar on hierarchical global planning for contact-rich manipulation using a graph of reachable sets.
10+
Invited talk at Duke Manipulation Seminar on hierarchical global planning for contact-rich manipulation using a graph of reachable sets.

‎files/Simin_Liu_CV_04_24_26.pdf‎

127 KB
Binary file not shown.

‎images/grs_query_115.gif‎

708 KB
Loading

‎images/grs_query_153.gif‎

2.14 MB
Loading

‎images/grs_query_161.gif‎

567 KB
Loading

‎images/grs_query_190.gif‎

791 KB
Loading

‎images/grs_query_201.gif‎

352 KB
Loading

0 commit comments

Comments
 (0)