Module register

Three pillars, compulsory at every level

A security specialist still commissions robots, and a robotics engineer still runs a live blue-team rotation. Hours are timetabled contact hours and exclude assessment and self-directed study.

Pillar A · AAI

Applied AI & Foundation Models

30% of core hours

The intelligence layer: how embodied policies are learned, evaluated, governed and kept honest. Deliberately light on architecture trivia and heavy on evaluation — the industry failure mode is not a badly chosen model, it is a model nobody could prove anything about.

Code	Module	Lvl	Hrs	Core content
AAI-101	Foundations of Machine Intelligence	L1	40	What learned systems can and cannot do; failure taxonomies; why a confident model is not a correct one
AAI-110	Data Literacy for Physical Systems	L1	30	Sensor data, sampling, drift, labelling; reading a performance envelope and a confidence interval
AAI-201	Perception Models for Robotics	L2	60	Detection, segmentation, pose estimation, depth; occlusion, reflectivity and adversarial conditions
AAI-210	Sensor Fusion & State Estimation	L2	50	Filtering, SLAM, multi-modal fusion; degradation behaviour when a sensor lies rather than fails
AAI-220	Applied ML Engineering & Tooling	L2	45	Training pipelines, versioning, experiment tracking, reproducibility; MLOps for embodied systems
AAI-301	Imitation Learning from Demonstration	L3	70	Behaviour cloning, DAgger, demonstration quality, covariate shift; designing a capture protocol
AAI-310	Reinforcement Learning at Scale	L3	70	GPU-parallel RL, reward design and reward hacking, curriculum learning, sim budget economics
AAI-320	Vision-Language-Action Models	L3	60	Generalist embodied policies, instruction following, grounding, task decomposition, prompt-level risk
AAI-330	Evaluation, Benchmarking & Statistical Envelopes	L3	50	Designing an evaluation that would fail if the policy were bad; sample size; tail behaviour
AAI-340	Data Engineering for Fleet Learning	L3	45	Trajectory stores, lineage, privacy and IP boundaries, the failure library, dataset governance
AAI-401	Multi-Fleet Policy Architecture	L4	60	Policy versioning across heterogeneous fleets, staged rollout, rollback, A/B under safety constraint
AAI-410	Model Risk & AI Governance	L4	40	NIST AI Risk Management Framework, model cards, accountability chains, regulatory landscape
AAI-420	Continual & On-Fleet Learning	L4	50	Learning after deployment without silently invalidating the certificate; drift detection
AAI-501	Evidence Standards for Learned Behaviour	L5	45	What constitutes admissible evidence that a learned policy is safe; argument structure; fallacies

Pillar B · CPS

Cyber-Physical Security

27% of core hours

Autonomy is an attack surface with actuators. IT security curricula stop at the data; robotics curricula never start on the adversary. Every module is taught in Zone 7, the cyber range, against equipment that moves.

Code	Module	Lvl	Hrs	Core content
CPS-101	Security Fundamentals for Machine Operators	L1	30	Credentials, media control, phishing in an OT context, why a USB stick is a mechanical hazard
CPS-110	Safe Stop, Isolation & Incident Reporting	L1	20	Recognising anomalous machine behaviour; isolation procedure; escalation without freezing
CPS-201	OT Networks & Industrial Protocols	L2	55	Modbus, EtherNet/IP, PROFINET, OPC UA, DDS and ROS transports; what is authenticated and what is not
CPS-210	IEC 62443 Zones, Conduits & Architecture	L2	45	Segmentation, conduits, security levels, the Purdue model and where robot fleets break it
CPS-220	Safety Instrumented Systems & Interlocks	L2	45	SIS design, independence, defeat detection, keeping safety functions out of the learning stack
CPS-301	Robot & Autonomy Attack Surface	L3	65	Firmware, controllers, ROS graph, sensor spoofing (LiDAR, GNSS, camera), actuator command injection
CPS-310	Secure Teleoperation & Command Integrity	L3	50	Latency as a weapon, session hijack, command authentication, dead-man design, degraded-link doctrine
CPS-320	Detection Engineering for Moving Machinery	L3	55	Physics-informed detection, process vs network anomaly, alerting a human can act on in 4 seconds
CPS-330	Adversarial Machine Learning	L3	50	Evasion, poisoning, backdoors in demonstration data, model extraction, defending fleet learning
CPS-340	Blue Team Rotation — Live Range	L3	60	Full rotation defending an operating bay under live attack, with production KPIs still measured
CPS-401	Red Team Rotation — Cyber-Physical	L4	70	Authorised offensive rotation with safety-observer discipline; reporting to a defensible standard
CPS-410	Supply Chain & Firmware Assurance	L4	45	SBOM for robots, update integrity, vendor assessment, secure boot and attestation
CPS-420	Incident Command with Energised Assets	L4	45	Command structure when the compromised asset weighs 400 kg; evidence preservation vs safety
CPS-501	Threat-Informed Safety Cases	L5	45	Integrating adversarial scenarios into the safety argument; residual risk under intelligent opposition

Pillar C · PAI

Physical AI & Embodied Systems

43% of core hours

The metal: control, mechanics, simulation, fleet operations and the functional-safety practice that makes autonomy deployable. It carries the largest share of hours because it carries the largest share of the ways a deployment fails.

Code	Module	Lvl	Hrs	Core content
PAI-101	Robot Awareness & Shared-Space Safety	L1	35	Working near autonomous machines; separation distances; predicting robot intent; emergency behaviour
PAI-110	Fleet Operations & Supervisory Control	L1	55	Fleet dashboards, task dispatch, exception handling, shift handover, when to intervene and when not to
PAI-120	Teleoperation Basics & Demonstration Capture	L1	40	Rig operation, latency compensation, capture protocol discipline, quality of a demonstration
PAI-201	Kinematics, Dynamics & Control	L2	60	Forward/inverse kinematics, trajectory generation, impedance and force control, stability
PAI-210	Actuation, Power & Thermal Systems	L2	50	Actuator types, transmissions, battery systems and thermal limits; duty cycle and derating
PAI-220	Calibration & Metrology	L2	45	Hand-eye calibration, extrinsics, repeatability vs accuracy, drift measurement
PAI-230	Diagnostics, Teardown & Return to Service	L2	65	Fault isolation, teardown and rebuild, component-level repair, documented return-to-service authority
PAI-240	Manipulation & End Effector Systems	L2	50	Grasp planning, compliant and soft grippers, tool changers, contact-rich task design
PAI-301	Locomotion & Whole-Body Control	L3	65	Bipedal and quadrupedal gait, balance recovery, terrain adaptation, whole-body manipulation
PAI-310	Simulation, Digital Twins & Domain Randomisation	L3	70	Building a twin worth trusting; physics fidelity, sensor models, randomisation coverage
PAI-320	Sim-to-Real Transfer Engineering	L3	60	Measuring the reality gap, transfer loss quantification, targeted twin refinement, transfer forensics
PAI-330	Fleet Orchestration & Traffic Management	L3	55	Multi-agent scheduling, deadlock, congestion, mixed human-robot traffic, degraded-fleet operation
PAI-340	Shared Autonomy & Human-Robot Handover	L3	50	Authority transfer, mode confusion, situation awareness after handover, designing the intervention
PAI-350	Functional Safety: ISO 10218 & ISO/TS 15066	L3	50	Risk assessment, speed and separation monitoring, power and force limiting, validation
PAI-401	Multi-Site Autonomous Operations Architecture	L4	65	Fleet architecture across sites, edge/cloud split, remote operations centres, standardisation vs local fit
PAI-410	Reliability, Availability & Fleet Economics	L4	45	MTBF/MTTR for fleets, spares strategy, availability modelling, real cost per autonomous task hour
PAI-420	Commissioning & Brownfield Integration	L4	55	Integrating autonomy into a plant that already works; cutover, interim states, stakeholders
PAI-501	Assurance Case Authoring & Witness Testing	L5	70	Structured argument, evidence sufficiency, designing and witnessing acceptance tests
PAI-510	Examiner Practice & Calibration of Judgement	L5	55	Examining without coaching, inter-examiner calibration, defensible failure decisions, appeals