Families / OPT

Optimisation & Planning

Finding the best decision under constraints — routing, scheduling, pricing, resource allocation (often ML forecast + solver).

Show its 41 use cases in the catalog

Intro & scope

Use cases where a mathematical optimiser (LP/MIP/CP-SAT, metaheuristics, RL policies) computes a best feasible decision under constraints; ML (forecasts, learned costs) feeds it and an LLM at most wraps it (modelling copilot, explanations). Patterns: OPT.LOGISTICS (routing, inventory, packing, network), OPT.SCHEDULE (production, workforce, timetables, dispatch), OPT.RESOURCE (energy, pricing, portfolio, allocation), OPT.PROCESS (plant/process parameters, control).

Scope boundaries: pure forecasting belongs to PRED (OPT consumes its output); anomaly alerting to ANOM; LLM-only agents to AGENT. Overlaps: OPT-001/019/028 (routing vs technician dispatch vs ridepooling: vehicle routing with fixed fleet vs skill/appointment-driven vs real-time on-demand), OPT-011/021/032/022/039 (battery, EV depot, home, grid operator, trading: same dispatch core, different asset owner and market role), OPT-009/038/024 (parameter tuning offline vs closed-loop control vs water-specific), OPT-003/035 (shift planning vs contact-centre routing; both employee-affecting, AI Act 'depends' via Annex III employment when worker allocation/monitoring is automated). Safety-critical OPT-029 is high-risk (emergency dispatch, Annex III 5(d)).

AI Act reasoning (common): minimal when decisions concern assets/goods; 'depends' when tasks/shifts are allocated to workers, or the system is a safety component of critical infrastructure (energy, water, traffic, Annex III 2); high where emergency services are triaged.

Use cases

ID Title Pattern Maturity Path HW tier AI Act risk
OPT-001 Vehicle routing & tour optimisation OPT.LOGISTICS mainstream A HW1 minimal
OPT-002 Production scheduling & advanced planning (APS) OPT.SCHEDULE mainstream A HW1 minimal
OPT-003 Workforce scheduling & shift planning OPT.SCHEDULE mainstream A HW1 depends
OPT-004 Dynamic pricing & revenue management OPT.RESOURCE mainstream A HW1 minimal
OPT-005 Inventory optimisation & automated replenishment OPT.LOGISTICS mainstream A HW1 minimal
OPT-006 Warehouse slotting & picking optimisation OPT.LOGISTICS mainstream A HW1 minimal
OPT-007 Building energy management & HVAC optimisation OPT.RESOURCE early-adopter A HW1 minimal
OPT-008 Supply chain network design & S&OP planning OPT.LOGISTICS mainstream A HW1 minimal
OPT-009 Manufacturing process parameter optimisation OPT.PROCESS early-adopter A HW1 depends
OPT-010 Portfolio optimisation & algorithmic trading OPT.RESOURCE mainstream A HW1 minimal
OPT-011 Battery storage dispatch & grid flexibility optimisation OPT.RESOURCE early-adopter A HW1 depends
OPT-012 Load, container & cutting optimisation (packing/nesting) OPT.LOGISTICS mainstream A HW1 minimal
OPT-013 Timetabling (schools, universities, public transport) OPT.SCHEDULE mainstream A HW1 minimal
OPT-014 Precision farming: irrigation & fertilisation optimisation OPT.PROCESS mainstream A HW1 minimal
OPT-015 Cloud cost & compute resource optimisation (FinOps) OPT.RESOURCE mainstream A HW1 minimal
OPT-016 Trade promotion optimisation & promotion uplift planning (CPG / retail) OPT.RESOURCE mainstream A HW1 minimal
OPT-017 Assortment planning & store space optimisation OPT.RESOURCE mainstream A HW1 minimal
OPT-018 Natural-language optimisation modelling & decision copilot OPT.PROCESS early-adopter B HW1 minimal
OPT-019 Field service scheduling & technician dispatch optimisation OPT.SCHEDULE mainstream A HW1 depends
OPT-020 Disruption management & operations recovery (airlines, rail, public transport) OPT.SCHEDULE early-adopter A HW1 depends
OPT-021 EV fleet & depot smart-charging optimisation OPT.RESOURCE early-adopter A HW1 minimal
OPT-022 Distribution grid congestion management & redispatch optimisation (DSO/TSO) OPT.RESOURCE early-adopter A HW1 depends
OPT-023 District heating & CHP dispatch optimisation OPT.RESOURCE early-adopter A HW1 depends
OPT-024 Water & wastewater treatment optimisation (aeration, dosing, pump scheduling) OPT.PROCESS early-adopter A HW1 depends
OPT-025 Adaptive traffic signal control OPT.SCHEDULE early-adopter A HW1 depends
OPT-026 Voyage & flight trajectory optimisation (weather routing, fuel, emissions, contrails) OPT.LOGISTICS mainstream A HW1 minimal
OPT-027 Construction & capital project schedule optimisation and delay-risk forecasting OPT.SCHEDULE early-adopter A HW1 minimal
OPT-028 On-demand ridepooling & demand-responsive transport dispatch OPT.LOGISTICS early-adopter A HW1 minimal
OPT-029 Emergency services resource deployment (ambulance dispatch, relocation & station planning) OPT.RESOURCE early-adopter B HW1 high
OPT-030 Mobile network (RAN) optimisation & energy saving OPT.RESOURCE early-adopter A HW1 depends
OPT-031 Recipe, blend & formulation optimisation (feed, food, concrete, fuels) OPT.PROCESS mainstream A HW1 minimal
OPT-032 Home energy management for prosumers (PV, battery, heat pump, EV charging on dynamic tariffs) OPT.RESOURCE early-adopter A HW1 minimal
OPT-033 Digital freight matching & spot-rate brokerage automation OPT.LOGISTICS early-adopter A HW2 minimal
OPT-034 Sales territory design, account assignment & quota setting OPT.RESOURCE mainstream A HW1 depends
OPT-035 Predictive customer-to-agent routing in contact centres OPT.RESOURCE early-adopter A HW1 depends
OPT-036 Fibre (FTTH) & utility network rollout planning and design automation OPT.RESOURCE early-adopter A HW2 minimal
OPT-037 Container terminal & port operations optimisation (berth, quay crane, yard stacking, truck slots) OPT.LOGISTICS early-adopter A HW1 depends
OPT-038 Closed-loop AI process control & real-time optimisation for continuous plants OPT.PROCESS early-adopter A HW1 depends
OPT-039 Renewable portfolio marketing & algorithmic power trading OPT.RESOURCE mainstream A HW1 depends

Shared stack

Common core (all tiers): solver layer (OR-Tools CP-SAT/routing, HiGHS, optionally Gurobi/CPLEX/Timefold), Pyomo/PuLP or linopy modelling, Postgres/TimescaleDB for data and results, Python orchestration (Dagster/Airflow/Prefect), LightGBM/Prophet/statsforecast for the predictive inputs, MLflow for model registry, a what-if UI (Streamlit/Superset), human approval step with audit log. EU-hosted (or on-prem) data; no personal data leaves the EU.

HW1 (CPU only, 8-32 cores, 32-128 GB RAM, one VM or small k8s node; nearly all OPT cases): the core above plus Valhalla/OSRM for road matrices, Ray/Optuna for parallel search, Stable-Baselines3 on CPU for small RL policies, SimPy/Mesa for simulation, MQTT/Kafka + Redis for near-real-time telemetry (energy, traffic, RAN). Shared by OPT-001/005/006/012/019/028 (routing + geodata), OPT-007/011/021/022/023/032/039 (energy: forecasts + MILP dispatch + MQTT/OpenADR/IEC 61850 gateways), OPT-009/024/038 (OPC UA historian + soft-sensor + MPC/RL).

HW2 (single GPU 24-48 GB, e.g. L40S/A6000): add a self-hosted open-weight LLM (Mistral/Llama/Qwen 7-32B via vLLM) for NL modelling copilot (OPT-018), document/voice intake (OPT-033), plus segmentation/vision pipelines (OPT-036); GPU RL training for large simulators.

HW3 (multi-GPU node): only for large foundation-model fine-tuning or massive RL training; not required by any OPT case today.

Cross-family reuse: forecasting inputs from PRED (same TimescaleDB + MLflow), anomaly gates from ANOM, document extraction from DOC, agent frameworks from AGENT (OPT-018 copilot generates solver code, executes in sandbox, and the solver stays the decision authority).