Chapter 17. Domain Playbook Autonomous Vehicles
Section 1. Operational Goals and Hazards
- Plan, localize, and navigate urban/suburban roads with bounded risk
- Handle edge cases: occlusions, jaywalkers, cut-ins, and rare weather
- Meet safety envelopes: stopping distance, TTC, and fallback behaviors
Section 2. Sensor Suite and Limits
- Camera/LiDAR/Radar/IMU/GNSS with redundancy and thermal constraints
- Failure modes: saturation, blooming, multipath, wheel slip
- Calibration persistence across shocks, temperature, and service cycles
Section 3. Fusion Stack (L0–L4)
- L0–L1: ego-motion + object tracks with lane/curb geometry fusion
- L2: scene graph (lanes, traffic controls, agents, right-of-way)
- L3: intent/interaction modeling (yield/merge/over-take)
- L4: sensor tasking, model selection, and map refresh policies
Section 4. Cognition Loop
- Goal hierarchy: navigation → path → behavior → actuation
- Attention to occlusion zones and vulnerable road users
- Rule arbitration: traffic code vs learned priorities
Section 5. Validation and Metrics
- Disengagements, infractions, comfort, and rare-event recall
- Closed-course, sim-at-scale, and shadow-mode comparisons
- Safety cases with traceability and evidence catalogs
Section 6. Incidents and Failure Modes
- Perception dropouts, stale maps, and misclassified signals
- Interaction deadlocks and unmodeled agent behaviors
- Recovery trees, minimum risk maneuvers, and handover
Section 7. Mini Case
- Unprotected left turn with pedestrian emergence at dusk
Chapter 18. Domain Playbook Smart Homes and Assistive Robotics
Section 1. Operational Goals and Hazards
- Reliable assistance (fetch, clean, reminders) in cluttered homes
- Privacy preservation and safe human–robot proximity
- Non-stationary layouts, pets/children, and device idiosyncrasies
Section 2. Sensor Suite and Limits
- RGB-D, tactile/force, microphone arrays, BLE/IoT beacons
- Limits: specular floors, soft-body grasping, reverberant speech
- Routine self-check calibration and auto-recovery prompts
Section 3. Fusion Stack (L0–L4)
- L0–L1: mapping/SLAM + graspable-object tracking
- L2: room/affordance semantics and activity contexts
- L3: task plans with user-preference priors
- L4: policy switching by time-of-day, resident, and energy budget
Section 4. Cognition Loop
- Goal inference from routines and lightweight dialogue acts
- Memory for object placement and user-specific constraints
- Social compliance and interruption handling
Section 5. Validation and Metrics
- Task success, time-to-complete, and assist burden reduction
- Speech-under-noise and grasp success under variability
- Privacy/consent audits and local-only processing rates
Section 6. Incidents and Failure Modes
- Confusable commands, reflective floors, and fragile objects
- Pet/child interference and blocked charge docks
- Safe-stop envelopes and user notification pathways
Section 7. Mini Case
- “Find my glasses” with room-level search and dialogue refinement
Chapter 19. Domain Playbook C2 / ISR and Defense
Section 1. Operational Goals and Hazards
- Persistent awareness, target development, and time-critical decisions
- Deception, spoofing, and contested/denied environments
- Rules of engagement, collateral constraints, and escalation control
Section 2. Sensor Suite and Limits
- EO/IR, SAR/GMTI, EW/ESM, AIS/ADS-B, HUMINT/OSINT
- Limits: comms intermittency, look-angle restrictions, weather
- Cross-platform alignment and latency-aware dissemination
Section 3. Fusion Stack (L0–L4)
- L0–L1: multi-INT track/ID with emitter correlation
- L2: order of battle, formations, and activity patterns
- L3: intent/threat and course-of-action hypotheses
- L4: sensor/asset tasking under bandwidth and risk budgets
Section 4. Cognition Loop
- Hypothesis management with evidence weighting and counterfactuals
- Human-on-the-loop approvals and red/blue/adversarial models
- Confidence-aware alerts and rationale summaries
Section 5. Validation and Metrics
- Track purity, latency-to-cue, and mission outcome uplift
- Robustness to deception and sensor loss
- Audit logs, chain-of-custody, and reproducible rationale
Section 6. Incidents and Failure Modes
- Identity swaps, decoys, and clutter-induced escalations
- Broken dissemination paths and stale intel fusion
- Graceful degradation playbooks and hold-fire safeguards
Section 7. Mini Case
- Maritime rendezvous with AIS spoofing and SAR revisit gaps
Chapter 20. Domain Playbook Healthcare and Clinical AI
Section 1. Operational Goals and Hazards
- Early detection, triage support, and workflow efficiency
- High stakes: false negatives, alarm fatigue, and bias risks
- Compliance, consent, and PHI minimization
Section 2. Sensor Suite and Limits
- Vitals, wearables, imaging, labs, EHR streams, clinician notes
- Limits: heterogeneity, missingness, drift from practice changes
- Device calibration, timestamp fidelity, and identity linkage
Section 3. Fusion Stack (L0–L4)
- L0–L1: signal cleaning + patient-state estimation
- L2: condition/syndrome hypotheses over multimodal evidence
- L3: risk/impact (deterioration, readmission, complications)
- L4: alert thresholds, cohort selection, and active testing
Section 4. Cognition Loop
- Clinician-in-the-loop decision support with explanations
- Preference-aware plans, contraindications, and shared decision making
- Uncertainty displays and “why not alerted?” rationale
Section 5. Validation and Metrics
- AUROC/PPV at alert volumes clinicians can sustain
- Calibration, subgroup fairness, and shift-resilience
- Silent trials, phased rollouts, and post-market surveillance
Section 6. Incidents and Failure Modes
- Data linkage errors, device drift, and guideline changes
- Alert floods and desensitization; automation bias
- Overrides, second-look policies, and escalation paths
Section 7. Mini Case
- Sepsis early warning with labs, vitals, and nursing notes
Chapter 21. Domain Playbook Industrial and Smart Cities
Section 1. Operational Goals and Hazards
- Throughput, yield, and OEE improvement with bounded downtime windows
- Worker safety, lockout/tagout integrity, and proximity risks
- Utility reliability, demand spikes, and cascading failures across assets
Section 2. Sensor Suite and Limits
- Vibration, thermography, power quality, PLC/SCADA tags, CCTV, traffic/air quality
- Limits: sensor drift, EMI, occlusions, seasonal cycles, policy redactions
- Asset identity linkage across CMMS, BIM/GIS, and telemetry namespaces
Section 3. Fusion Stack (L0–L4)
- L0–L1: condition indicators + object/event tracks for assets, people, vehicles
- L2: site/city situation graphs (zones, flows, hazards, emissions)
- L3: impact on safety, uptime, SLAs, and environmental compliance
- L4: tasking sensors, rerouting flows, and dynamic work orders
Section 4. Cognition Loop
- Multi-objective planning (safety → production → energy → emissions)
- Preference/priority models per shift, zone, and regulation
- Coordination with human operators and automated substations
Section 5. Validation and Metrics
- MTBF uplift, downtime reduction, false alarm rate, and maintenance precision
- Traffic KPIs: delay, queue length, emergency preemption success
- Environmental KPIs: exceedance detection, time-to-mitigation
Section 6. Incidents and Failure Modes
- Sensor dropouts, PLC tag changes, data skew, and silent unit swaps
- Conflicting objectives (throughput vs safety vs energy)
- Playbooks for safe fallback, manual override, and audit capture
Section 7. Mini Case
- Heatwave event: load shedding + traffic retiming + HVAC fault isolation