Tutorials#
Step-by-step tutorials that start from the basics and then work through the models one by one. Each
tutorial is a notebook that runs top to bottom in a few seconds to a minute; the pages below were
executed when this documentation was built, so every number and figure is real output. The sources
are Markdown files in tutorials/
and can be converted to notebooks with python tutorials/_convert.py.
Simulated data are labelled as simulated, and every model is compared with a classical baseline. Where the classical method wins, the tutorial says so.
flowchart LR
accTitle: Reading order of the tutorials
accDescr: The introduction leads to four tracks: models of people (2, 3), the robot decision layer (4 to 7, then 12, 13 and the packaged hand-over use case 14), perception (8, 9) and the COBRA belief filter and viewer (10, 11), and the quantum computing and quantum sensing threads (15, 16).
T1["1 · Introduction"]:::a --> T2["2 · Question order"]:::h --> T3["3 · Trust on a qubit"]:::h
T1 --> T4["4 · Ask or act"]:::r --> T5["5 · Calibration"]:::r --> T6["6 · Orchestration"]:::r --> T7["7 · Question planner"]:::r
T5 --> T8["8 · Fusion"]:::p --> T9["9 · Bistable perception"]:::p
T1 --> T10["10 · COBRA belief filter"]:::q --> T11["11 · Bloch sphere"]:::q
T1 --> T15["15 · Quantum computing for robots"]:::q
T1 --> T16["16 · Quantum sensing"]:::q
T7 --> T12["12 · End to end"]:::r
T8 --> T12
T3 --> T12 --> T13["13 · ROS 2"]:::r --> T14["14 · Hand-over package"]:::r
classDef a fill:#0b2a4a,stroke:#79c0ff,color:#e6edf3
classDef h fill:#3d1414,stroke:#ff7b72,color:#e6edf3
classDef r fill:#2a1b3d,stroke:#d2a8ff,color:#e6edf3
classDef p fill:#3a2a0a,stroke:#ffa657,color:#e6edf3
classDef q fill:#1f2a3a,stroke:#79c0ff,color:#e6edf3,stroke-dasharray:4 3
Start here#
Beliefs as vectors, questions as projectors, why order matters, and the fit-and-compare workflow.
Models of people#
Fit the Clinton-Gore poll, run the QQ test, check model recovery.
Trust on the Bloch sphere, the effect of asking, and a trust-aware hand-over policy.
Robot decision layer#
Expected cost, model disagreement and hazard-aware decisions.
Reliability diagrams, Platt, temperature and isotonic recalibration, coverage guarantees.
A cost-aware gate and a meta-learned prior for new tasks.
Value of information with order-dependent answers.
Calibration, fusion, questions and trust-aware hand-over in one robot loop.
The JSON service and the ROS 2 node, run without a ROS installation.
An installable package with a benchmark, a PyBullet arm simulation (GIF), a ROS 2 node and tests.
Perception#
Adapters, Bayesian, Dempster-Shafer and quantum-like fusion.
The quantum Zeno model against Markov and gamma-renewal baselines.
COBRA belief filter and the viewer#
A density-matrix belief whose measurement basis depends on the previous frame, against a Bayes filter.
Trajectories, animations and tomography for circuits and for the trust qubit.
Quantum computing and quantum sensing for robots#
Robot problems as QUBOs, solved classically and with QAOA, against a random-sampling null baseline.
Quantum limits, an atom-interferometer accelerometer and a Kalman filter that fuses it with a MEMS sensor.