Why focus on the chess route first?
Because it is a concrete, testable benchmark that can close the full loop on a tabletop. It is not the end state, but a transition route toward household robotics.
In DevelopmentV1.01
SOMA Arm
The most concrete reference implementation today
A tabletop language-driven robot arm project used to validate ANIMA's cognition loop, visual grounding, skill execution, and failure recovery.
It is not the final product. It is the benchmark and manipulation capability layer on the path to home robotics.
The real workstation, arm, and chessboard form the clearest system carrier right now.
Demo
Demo Video
V1.01 Demo — Latest progress (perception-foundation stage)
Hardware
Arm
Waveshare RoArm-M2-S 4-DOF
Camera
A Logitech C922 in an overhead tabletop setup
Workbench
Chessboard, chess pieces, and fixed lighting
Software
Cognition
ANIMA + Claude parser + py_trees
Perception
Grounding DINO + SAM2
Execution
ROS 2 + MoveIt2 + ACT / LeRobot
Current scope
- Identify capturable chess pieces on the board, understand chess rules, and execute capture moves
- Visual perception and board state recognition
- Failure detection, retry, and natural-language feedback
Working now
- Camera parameter lock, intrinsic calibration, eye-to-hand calibration, and world coordinate grounding
- Workspace reachability validation and perception pipeline setup
Future scope
- Board state recognition and chess engine integration
- Migration from a fixed tabletop arm to a fuller home robot platform
Milestones
Version Roadmap
2026-04-07 ~ 2026-04-13
V1.00 — Dev Environment & Basic Control
WSL2 + ROS 2 environment setup, RoArm-M2-S serial driver, MoveIt2 bridge, PDP Xbox gamepad teleop, C922 camera TCP bridge.
2026-04-14 ~
V1.01 — Perception Foundation
Camera parameter lock, intrinsic calibration, eye-to-hand calibration, world coordinate grounding, workspace reachability validation.
Planned
V1.02P — Chess Piece Manipulation
MoveIt2 primitives, Teleop data collection, ACT training.
Planned
V1.03P — Board Perception & Game Engine
Board state recognition, engine framework (in ANIMA), chess engine.
Planned
V1.04P — ANIMA Integration & Demo
Behavior tree orchestration, capture flow, NL frontend, demo video.
Public links
Demo / Media
Demo reel placeholder
Public demo video entry will be added later
FAQ
Frequently Asked Questions
Which parts are open source?
The main open tracks today are ANIMA and soma-arm. More future-facing productized lines may move into private repos, but this site will always label that status explicitly instead of pretending future plans are already public.
What exactly is the relationship between SOMA and ANIMA?
SOMA is the robot body and product family. ANIMA is the shared cognition framework. The homepage leads with SOMA because visitors understand a concrete robot before they understand the underlying framework.