From 6130f61821db0aa426d77948366e12195a9c0530 Mon Sep 17 00:00:00 2001 From: Imggaggu Date: Mon, 18 Aug 2025 23:57:43 +0900 Subject: [PATCH 1/5] [Fix] footer & title --- index.html | 75 ++++++++++++++++-------------------- process.html | 11 ++++++ static/images/favicon.svg | 37 ------------------ static/images/ticklebot.png | Bin 0 -> 1563001 bytes 4 files changed, 45 insertions(+), 78 deletions(-) create mode 100644 process.html delete mode 100644 static/images/favicon.svg create mode 100644 static/images/ticklebot.png diff --git a/index.html b/index.html index 373119fe36..7d568f101b 100644 --- a/index.html +++ b/index.html @@ -3,10 +3,10 @@ - + content="Low-cost humanoid control platform with integrated AI-ROS2 system for daily task automation"> + - Nerfies: Deformable Neural Radiance Fields + Ticklebot: Low-cost humanoid control platform with integrated AI-ROS2 system for daily task automation @@ -32,7 +32,7 @@ - + @@ -52,29 +52,31 @@
- +
- More Research + Learn More
- - HyperNeRF + + Process - - Nerfies + + Code +
@@ -88,34 +90,33 @@
-

Nerfies: Deformable Neural Radiance Fields

+

Ticklebot: Low-cost humanoid control platform with integrated AI-ROS2 system for daily task automation

- Keunhong Park1, - - Utkarsh Sinha2, + Dongje lee*, - Jonathan T. Barron2, - + HeeJae Chung*, - Sofien Bouaziz2, + Suhjung Park*, - Dan B Goldman2, + Wook Ko*, - Steven M. Seitz1,2, - - - Ricardo Martin-Brualla2 + Yunchang Choi* +

+
+ * Equal contribution +

+
- +
@@ -440,7 +441,7 @@

Related Links

- +
@@ -184,11 +192,11 @@

TickleBot Logo

- ROS2–AI Integrated Dual-Arm Control with SmolVLA + ROS2–AI Integrated Dual-Arm Control with various robot models for Physical AI

- TickleBot employs two low-cost SO-100 open-source robotic arms integrated - through ROS2 middleware and the SmolVLA vision-language-action model. + through ROS2 middleware and various robot models for Physical AI, such as SmolVLA, ACT policy, and NoMaD.
- This enables precise manipulation tasks such as folding and sorting laundry, @@ -212,10 +220,10 @@

TickleBot Logo

- NoMaD-Based Autonomous Navigation and Task Execution + NoMaD-Based Autonomous Navigation and exploration with topomap generation

- - The mobile base of TickleBot leverages the NoMaD model + - The mobile base of TickleBot leverages the NoMaD model to achieve vision-based navigation using only RGB cameras.
@@ -296,7 +304,7 @@

Abstract

- This project introduces TickleBot, a low-cost open-source robotic research platform that lowers the entry barrier to humanoid robotics for students and researchers. + This project introduces TickleBot, a low-cost open-source robotic research platform that lowers the entry barrier to Physical AI research for students and researchers.

Recent advances in Physical AI have highlighted the importance of foundation models for robotics, @@ -304,6 +312,7 @@

Abstract

continue to hinder progress and scalability. In this work, we present TickleBot, a low-cost, open-source robotic research platform designed to lower the entry barrier for undergraduate students and researchers. Built at approximately USD 725, TickleBot integrates the SO-100 open-source robotic arm with a ROS2-based middleware, enabling precise manipulation and mobility without requiring a physical twin. By incorporating the NoMaD model for mobile base navigation and SmolVLA for upper-body control, the platform supports AI-driven teleoperation and task execution using only RGB camera inputs. + Owing to the aforementioned characteristics, Ticklebot can attract a large pool of researchers, supporting high-quality teleoperation data collection and robot model evaluation, thereby contributing to the scaling of robot foundation models.

TickleBot is designed not only as a research tool @@ -445,7 +454,7 @@

Coding & ROS2 Integration

- On a laptop, ROS2 nodes and the SmolVLA model are executed, + On Jetson Orin nano(Jetpack 6.2.1), ROS2 nodes and the robot action model such as SmolVLA, ACT policy are executed, and the robotic arm performs basic motion tests to verify integration.

@@ -470,7 +479,7 @@

Hands-on Learning

- Students directly input commands or add new datasets, + Students directly input commands or add new datasets, via teleoperation system using joystick and robopilot handles. enabling the robotic arm to grasp and stack blocks as a practical learning exercise.

@@ -482,7 +491,7 @@

Research Extension

- The NoMaD-based mobile base navigates to a new workspace, where the robotic arm resumes manipulation tasks, + The NoMaD-based mobile base navigates to a new workspace(based on user input image) where the robotic arm resumes manipulation tasks, demonstrating scalability for advanced research.

@@ -505,13 +514,13 @@

Related Links

NoMaD: Navigation with Multimodal Diffusion A vision-based navigation model that enables robots to reach target locations using only RGB inputs.

- LeRobot An open-source library for robot learning that provides datasets, models, and training utilities for scalable robotic research. + LeRobot An open-source library for robot learning that provides datasets, models, and training utilities for scalable robotic research.

- SmolVLA A vision-based navigation model that enables robots to reach target locations using only RGB inputs. + SmolVLA A vision-based navigation model that enables robots to reach target locations using only RGB inputs.

- Mobile ALOHAA low-cost mobile humanoid platform combining dual arms with a wheeled base, demonstrating household task automation. + Mobile ALOHA A low-cost mobile humanoid platform combining dual arms with a wheeled base, demonstrating household task automation.