casual write-up about AI and Modular Robotics

Gathering some resources about Modular Robotics

cluster: tech, origami
(IoT, Robotics) อยากทำ write up ว่า modular/self essembled robots ที่ผ่านมามีการประยุกต์ใช้ ML หรือ NN แบบไหน ในการประกอบตัวเอง, เคลื่อนไหวประสานงานกัน:eyes:
inspiration:
Jamie paikทำ Lab ที่ RRL-EPFL

What is modular robots

from https://www.sciencedirect.com/science/article/abs/pii/S0094576521004938

• This term was introduced by Fukuda with the biologically-inspired CEllular roBOT (CEBOT) in the late 1980s.
  from https://www.robotpark.com/academy/all-types-of-robots/modular-robots/
• Each module has 2 or more connectors, and some also contain actuators to manipulate its location.
• it can be generally classified into several architectural groups by the geometric arrangement of their unit (lattice vs. chain). Several systems exhibit hybrid properties.

Why modular robots?

• space exploration

  • 
  • MIT News-Mix-and-match kit could enable astronauts to build a menagerie of lunar exploration bots

• disaster relief

  • its ability to change its own shape according to the environment for crawling, rolling, walking, and much more is suitable to this task.

• from https://www.robotpark.com/academy/all-types-of-robots/modular-robots/

  • ability to physically adapt, grow, heal, and even self replicate.

Interesting resources:

Modular Robotics: When You Want More Robots In Your Robot
https://hackaday.com/2018/05/10/modular-robotics-when-you-want-more-robots-in-your-robot/

Modular Robots: Mechanism and Control Design
https://www.ri.cmu.edu/project/modular-robots-mechanism-and-control-design/

• keywords: GNN, model-based reinforcement learning algorithms
• input a robot design, and outputs a controller for that design.
  • each module has DNN that get inputs from sensor measurements and output as actuator commands
  • messages passed from/to its neighbors (full robot sensors) assembling into GNN that output full robot commands

Learning Modular Robot Locomotion from Demonstrations(2022)
https://arxiv.org/abs/2210.17491
Keywords: Reinforcement and imitation learning(RL+IL)

• use RL for train control policy(p.2)
• IL for behavioral cloning

Trinity: A Modular Humanoid Robot AI System(2025)
https://arxiv.org/abs/2503.08338

• RL, LLM, and VLM
  • Use LLM to command robot’s movement (task planner)
  • RL for mobility
• not that modular, but it pique my interests.

TEDTalk about Robogami (some of them are modular)

Mori: A Modular Origami Robot
https://ieeexplore.ieee.org/document/7907345

• command it to move with quasi-2-D to 3-D configurations.
• what interesting about this paper: 100+ cites in this paper -> inspired and linked to many latest cool origami, robots, virtual reality projects.
  • e.g. Shiftly: A Novel Origami Shape-Shifting Haptic Device for Virtual Reality(2025)

created on: Mon Aug 25 2025