Merge branch 'master' of github.com:yukezhu/stanford-pair-website

README.md

@@ -2,4 +2,6 @@
 
 This repo contains the Jekyll files for the [Stanford People, AI & Robots (PAIR) Group website](https://pair.stanford.edu). Please submit pull requests when updating the website.
 
+To compile site for production, use `JEKYLL_ENV=production bundle exec jekyll build`.
+
 Maintainers: [Yuke Zhu](https://web.stanford.edu/~yukez/), [Andrey Kurenkov](http://www.andreykurenkov.com/), [Ajay Mandlekar](http://web.stanford.edu/~amandlek/)

_data/people.yml

@@ -142,6 +142,13 @@ current_students:
     email: qb1 [at] stanford [dot] edu
     link: //ameliabqy.github.io
 
+  tieler_callazo:
+    name: Tieler Callazo
+    img: tieler_callazo.jpg
+    title: Master's Student, ME
+    email: tieler [at] stanford [dot] edu
+    link: //www.linkedin.com/in/tielercallazo/
+
   jiren_zhu:
     name: Jiren Zhu
     img: jiren_zhu.png

index.md

@@ -35,6 +35,7 @@ about: "PAIR website"
       <div class="col-lg-12">
           <h2 class="page-header">News</h2>
           <ul class="news-list">
+            <li>We released <a href="https://sites.google.com/view/visionandtouch">our new paper</a> on learning multimodal representations for robot manipulation.</li>
             <li>Two papers from PAIR are accepted at <a href="http://www.robot-learning.org/">CoRL 2018</a>.</li>
             <li>We have released <a href="publications">our new paper</a> on one-shot visual imitation with neural task graphs.</li>
           </ul>
@@ -48,22 +49,35 @@ about: "PAIR website"
       <div class="col-lg-12">
           <h2 class="page-header">Project Highlights</h2>
       </div>
-      <div class="col-md-4">
-          <a href="https://arxiv.org/abs/1807.03480">
-            <img class="img-responsive img-hover" src="./img/project_thumbs/750x400/neural_task_graph.png" alt="">
+      <div class="col-md-4 mar-bot-25">
+          <a href="https://sites.google.com/view/visionandtouch">
+            <img class="img-responsive img-hover" src="./img/project_thumbs/750x400/multimodal.png" alt="">
           </a>
       </div>
-      <div class="col-md-4">
+      <div class="col-md-4 mar-bot-25">
+          <a href="http://roboturk.stanford.edu">
+            <img class="img-responsive img-hover" src="./img/project_thumbs/750x400/roboturk.png" alt="">
+          </a>
+      </div>
+      <div class="col-md-4 mar-bot-25">
+          <a href="http://surreal.stanford.edu">
+            <img class="img-responsive img-hover" src="./img/project_thumbs/750x400/surreal.png" alt="">
+          </a>
+      </div>
+      <div class="col-md-4 mar-bot-25">
           <a href="https://sites.google.com/view/task-oriented-grasp">
               <img class="img-responsive img-hover" src="./img/project_thumbs/750x400/task_oriented_grasping.png" alt="">
           </a>
       </div>
-      <div class="col-md-4">
+      <br>
+      <div class="col-md-4 mar-bot-25">
           <a href="https://stanfordvl.github.io/ntp/">
             <img class="img-responsive img-hover" src="./img/project_thumbs/750x400/neural_task_programming.png" alt="">
           </a>
       </div>
   </div>
+
+
   <!-- /.row -->
 
 

publications/_posts/2018-10-24-fan-corl18-surreal.md

@@ -0,0 +1,16 @@
+---
+layout: publication
+title: "SURREAL: Open-Source Reinforcement Learning Framework and Robot Manipulation Benchmark"
+authors: "Linxi Fan*, Yuke Zhu*, Jiren Zhu, Zihua Liu, Orien Zeng, Anchit Gupta, Joan Creus-Costa, Silvio Savarese, Li Fei-Fei"
+pub_info_name: "Conference on Robot Learning (CoRL)"
+pub_info_date: October 2018
+excerpt: text text text
+images:
+  thumb: fan-surreal-corl18.png
+  main: fan-surreal-corl18.png
+paper_link: "http://web.stanford.edu/~yukez/papers/corl2018surreal.pdf"
+webpage_link: "http://surreal.stanford.edu"
+video_link: "https://www.youtube.com/watch?v=efeS9AczbTk"
+code_link: "https://github.com/SurrealAI/Surreal"
+---
+Reproducibility has been a significant challenge in deep reinforcement learning and robotics research. Open-source frameworks and standardized benchmarks can serve an integral role in rigorous evaluation and reproducible research. We introduce SURREAL, an open-source scalable framework that supports stateof-the-art distributed reinforcement learning algorithms. We design a principled distributed learning formulation that accommodates both on-policy and off-policy learning. We demonstrate that SURREAL algorithms outperform existing open-source implementations in both agent performance and learning efficiency. We also introduce SURREAL Robotics Suite, an accessible set of benchmarking tasks in physical simulation for reproducible robot manipulation research. We provide extensive evaluations of SURREAL algorithms and establish strong baseline results.

publications/_posts/2018-10-24-mandlekar-corl18-roboturk.md

@@ -0,0 +1,15 @@
+---
+layout: publication
+title: "RoboTurk: A Crowdsourcing Platform for Robotic Skill Learning through Imitation"
+authors: "Ajay Mandlekar, Yuke Zhu, Animesh Garg, Jonathan Booher, Max Spero, Albert Tung, Julian Gao, John Emmons, Anchit Gupta, Emre Orbay, Silvio Savarese, Li Fei-Fei"
+pub_info_name: "Conference on Robot Learning (CoRL)"
+pub_info_date: October 2018
+excerpt: text text text
+images:
+  thumb: mandlekar-roboturk-corl18.png
+  main: mandlekar-roboturk-corl18.png
+paper_link: "http://vision.stanford.edu/pdf/mandlekar2018corl.pdf"
+webpage_link: "http://roboturk.stanford.edu"
+code_link: "https://github.com/StanfordVL/robosuite"
+---
+Imitation Learning has empowered recent advances in learning robotic manipulation tasks by addressing shortcomings of Reinforcement Learning such as exploration and reward specification. However, research in this area has been limited to modest-sized datasets due to the difficulty of collecting large quantities of task demonstrations through existing mechanisms. This work introduces ROBOTURK to address this challenge. ROBOTURK is a crowdsourcing platform for high quality 6-DoF trajectory based teleoperation through the use of widely available mobile devices (e.g. iPhone). We evaluate ROBOTURK on three manipulation tasks of varying timescales (15-120s) and observe that our user interface is statistically similar to special purpose hardware such as virtual reality controllers in terms of task completion times. Furthermore, we observe that poor network conditions, such as low bandwidth and high delay links, do not substantially affect the remote users' ability to perform task demonstrations successfully on ROBOTURK. Lastly, we demonstrate the efficacy of ROBOTURK through the collection of a pilot dataset; using ROBOTURK, we collected 137.5 hours of manipulation data from remote workers, amounting to over 2200 successful task demonstrations in 22 hours of total system usage. We show that the data obtained through ROBOTURK enables policy learning on multi-step manipulation tasks with sparse rewards and that using larger quantities of demonstrations during policy learning provides benefits in terms of both learning consistency and final performance.

publications/_posts/2018-10-25-lee-icra18-multimodal.md

@@ -0,0 +1,16 @@
+---
+layout: publication
+title: "Making Sense of Vision and Touch: Self-Supervised Learning of Multimodal Representations for 
+Contact-Rich Tasks"
+authors: "Michelle A. Lee*, Yuke Zhu*, Krishnan Srinivasan, Parth Shah, Silvio Savarese, Li Fei-Fei, Animesh Garg, Jeannette Bohg"
+pub_info_name: "arXiv preprint"
+pub_info_date: October 2018
+excerpt: text text text
+images:
+  thumb: lee-arxiv18-multimodal.png
+  main: lee-arxiv18-multimodal.png
+paper_link: "https://arxiv.org/abs/1810.10191"
+webpage_link: "https://sites.google.com/view/visionandtouch"
+video_link: "https://youtu.be/TjwDJ_R2204"
+---
+Contact-rich manipulation tasks in unstructured environments often require both haptic and visual feedback. However, it is non-trivial to manually design a robot controller that combines modalities with very different characteristics. While deep reinforcement learning has shown success in learning control policies for high-dimensional inputs, these algorithms are generally intractable to deploy on real robots due to sample complexity. We use self-supervision to learn a compact and multimodal representation of our sensory inputs, which can then be used to improve the sample efficiency of our policy learning. We evaluate our method on a peg insertion task, generalizing over different geometry, configurations, and clearances, while being robust to external perturbations. Results for simulated and real robot experiments are presented.