Perceive
Vision and sensors feed scene understanding into the World Model. Optional YOLO backend for pretrained detection.
A MODULAR VLA STACK
SIM-FIRST · MOCK + MUJOCO
Lightweight modular Vision–Language–Action for robotics. Not one giant transformer—specialized modules talk through a World Model.
Vision–Language–Action at the Intersection of Software & Robotics
World Model is the integration bus. Language never emits joint commands. Safety lives in the motion controller. LLM and planner run on events, not every tick.
THE PIPELINE
How it moves
Understand
Language and planner turn instructions into subtasks—event-driven, not every control tick. Optional BitNet for local reasoning.
Act
Policy → control → robot. BC-trained policy, damped-least-squares IK, and optional LeRobot/SmolVLA pretrained checkpoints.
Learn
train-bc for behavior cloning, multi-task eval across pick and place, and vla-bench for systems benchmarks.
A Stack You Can Run Today
BackendDetailEntryStatus
Mock simAlways works — lightweight arm, no extra depskinetic demo pick60 s
MuJoCoPhysics + 3D viewer tabletop scenekinetic demo mujocoOptional
BitNet LLM1.58-bit local reasoner on replan eventsconfigs/bitnet.yamlOptional
LeRobot / SmolVLAPretrained HF VLA checkpoints through the looppolicy.backend: lerobot~27 s / 50-action chunk on CPU
ROS2Joint trajectory adapter for real armsrobots.ros2Dry-run
See. Think. Move.
Benchmarks
Benchmarked against lightweight VLAs
Latency and literature comparison with open VLAs—honest about what Kinetic is and isn't. Systems efficiency on the control loop, plus a gaps roadmap for what still needs real robots, vision, and trained skills.
FIG. 02 — SYSTEMS LATENCY