Claude Academy ↗
Anthropic’s courses, tutorials and AI fluency resources.
フィジカルAI
New research: robot recovery, data interfaces and executable motion →
From coordinate frames to robot foundation models. A connected study and interview-preparation path with diagrams, equations, worked examples and evidence-based project checkpoints.
Technical posts currently have English and Chinese editions. This locale displays the English article with localized navigation.
These are original learning exercises, not company interview reports. Reading supports preparation; hiring depends on practical skills and role fit. Articles and diagrams work without uploads. Proposed robot-training experiments require your own simulation or hardware environment.
Choose a role, then build evidence
Robot Learning / VLA · 1 → 4 → 5 → 6 → 7 → 8. Deliver a dataset audit, baseline and ablation report.
Perception / Controls · 1 → 2 → 3 → 8. Deliver calibration, estimation and disturbance-response tests.
Robotics Infrastructure · 2 → 3 → 6 → 8. Deliver timestamp contracts, action adapters and latency/failure dashboards.
- 1
01 · Frames, transformations and robot kinematics
Translate a camera observation into a robot command without losing frame, unit or rotation conventions.
Prerequisites: Vectors, matrix multiplication, radians.
Read post & exercises → - 2
02 · Perception, state estimation and time
Understand why a sharp image can still produce a stale, uncertain control state.
Prerequisites: Frames, probability and covariance.
Read post & exercises → - 3
03 · Feedback control and receding-horizon planning
Connect learned actions to stable feedback and constrained execution.
Prerequisites: Dynamics, derivatives and linear algebra.
Read post & exercises → - 4
04 · Demonstrations, behavior cloning and ACT
Design a demonstration dataset and explain why action chunks help without hiding distribution shift.
Prerequisites: Supervised learning, sequence models and control frequency.
Read post & exercises → - 5
05 · Diffusion Policy and multimodal actions
Explain why averaging demonstrations can fail and how conditional denoising represents alternatives.
Prerequisites: Behavior cloning, Gaussian noise and sequence prediction.
Read post & exercises → - 6
06 · OpenVLA: vision, language and action interfaces
Trace a language-conditioned robot policy through action encoding, normalization and embodiment.
Prerequisites: Transformers, imitation learning and robot frames.
Read post & exercises → - 7
07 · π₀ and continuous action flow matching
Distinguish token prediction, diffusion noise prediction and a continuous velocity field.
Prerequisites: VLA interfaces, differential equations and conditional generation.
Read post & exercises → - 8
08 · Robot RL, sim-to-real and deployment evidence
Design rewards and evaluations that reveal failure before making claims about real robots.
Prerequisites: MDPs, PPO, control constraints and statistical evaluation.
Read post & exercises →
Interview readiness: demonstrate, do not memorize
For every topic: derive the key equation without notes, reproduce the numeric example, implement one bounded test, and explain a failure. Score each dimension 0 (not yet), 1 (with help), or 2 (independently). A high self-score is a practice signal, not a hiring prediction.
Continue with official AI education
OpenAI Academy ↗
Learning resources for builders, educators and AI workflows.
Microsoft Learn · AI ↗
Role-based AI learning paths and technical training.
External official resources, checked 25 October 2026. Independent site; no endorsement or affiliation. These links leave Token Lab and follow the destination’s privacy and sign-in policies.