Current Version: v0.5.1
PID steering with asymmetric stride, meta-learning loop (Phase A-D), hardware drift simulation, Freenove phototaxis validated on real hardware.
Near Term
- Meta-Learning validation — longer runs (100k+ steps) and harder scenarios to activate the autonomous strategy adaptation loop
- Hardware video — document PID steering and phototaxis on the real Freenove robot
- Third robot platform — port to Petoi Bittle X V2 to prove body-agnostic architecture
- Resume training — validate brain persistence across sessions with meta-learning state
Medium Term
- Baby-KI v0.8.0 — pure intrinsic reward (no external reward signal), autonomous learning through curiosity and prediction error
- Olfactory navigation — run-and-tumble chemotaxis with impulse-based orientation (biological E. coli strategy)
- Neuromorphic chip deployment — port SNN to Intel Loihi or SynSense Speck for sub-watt inference
- Conference submission — IROS, CoRL, or RSS paper with multi-platform results
Long Term
- On-hardware meta-learning — the real robot learns and adapts autonomously, not just replays sim-trained weights
- Multi-robot interaction — social behaviors between multiple MH-FLOCKE dogs
- Cortical layers — abstract planning, sequence learning, concept formation beyond the current cerebellar/brainstem level
Non-Goals
- LLM integration for motor control — language models are not motor controllers
- Scaling to millions of neurons without biological justification
- Cloud-dependent inference — MH-FLOCKE runs on-device, always