Operations, Logistics, Energy and Manufacturing

Introduction

Quantum computing is increasingly investigated for problems in which many possible decisions must be coordinated under constraints. Operations, logistics, energy and manufacturing contain precisely this type of problem: vehicles must visit locations efficiently, employees must be assigned to shifts, production must respect capacity and demand, and electrical systems must balance supply, storage and consumption in real time.

The practical importance of quantum computing in these industries does not arise because quantum computers automatically solve every optimization problem faster. Rather, quantum methods provide alternative representations and search strategies for difficult problems, especially combinatorial optimization problems. Current hardware remains limited by noise, scale and connectivity. Consequently, the most credible near-term approach is hybrid quantum–classical computing, in which classical systems prepare data, quantum processors explore structured alternatives, and classical algorithms evaluate and improve the results.

This chapter examines how quantum computing may support vehicle routing, workforce scheduling, production planning, grid optimization, energy systems, predictive maintenance and industrial digital transformation. It also evaluates implementation constraints, governance requirements and the difference between a promising experiment and a useful industrial solution.

Learning objectives

By the end of this chapter, the learner should be able to:

1. Explain why industrial planning problems are suitable candidates for quantum optimization.
2. Formulate routing, scheduling and allocation problems using objective functions and constraints.
3. Distinguish QAOA, quantum annealing, variational methods and classical heuristics at a practical level.
4. Identify applications in logistics, manufacturing, energy and predictive maintenance.
5. Evaluate quantum pilots using operational, financial and technical metrics.
6. Discuss data, safety, cybersecurity, ethical and governance issues in industrial deployment.

1. Industrial optimization as a decision problem

An industrial optimization problem consists of decisions, an objective and constraints. A decision variable records a choice, such as whether vehicle k travels from location i to location j. The objective measures performance, such as cost, distance, delay, energy consumption or carbon emissions. Constraints specify what is permitted, such as vehicle capacity, employee availability or power-balance requirements.

A general constrained optimization model can be written as:

Here, x is the vector of decisions, f(x) is the objective function, and gᵣ(x) and hₛ(x) are inequality and equality constraints. The formulation is applicable when decisions and restrictions can be expressed mathematically. In industrial problems, some variables are binary, meaning that they take values 0 or 1, while others are integer or continuous.

For example, a binary variable xᵢⱼ may equal 1 if a route travels directly from location i to location j, and 0 otherwise. A linear cost objective is:

where cᵢⱼ is the cost of travelling from i to j. If cᵢⱼ is measured in kilometres, C is in kilometres; if it is monetary cost, C is monetary units. The formula assumes that route costs are additive and known or estimated before optimization.

Many industrial models are NP-hard. This means that no general algorithm is known that solves every instance efficiently as problem size increases. NP-hardness does not mean that a problem is impossible; it means that exact solution may become computationally expensive. Classical mixed-integer programming, constraint programming, local search and metaheuristics remain highly effective. Quantum computing should therefore be assessed against strong classical baselines rather than against an unrealistic assumption that classical methods are incapable.

2. From industrial problem to quantum model

Quantum optimization requires a transformation from the operational model to a form compatible with a quantum algorithm. A common approach is binary quadratic optimization. Its objective is represented as:

where xᵢ ∈ {0,1}, aᵢ are individual decision coefficients and bᵢⱼ represent interactions between decisions. A linear constraint can be incorporated by a penalty. For example, if exactly one choice must be selected, the constraint is ∑ᵢ xᵢ = 1. A penalized objective is:

where P > 0 is a penalty coefficient. Feasible solutions incur zero penalty for this constraint, while violations incur a positive penalty. The penalty must be sufficiently large to discourage infeasible solutions, but excessively large values can make the objective numerically poorly balanced. Penalty selection is therefore a modelling and calibration task, not a purely quantum task.

Binary quadratic models can be mapped to Ising models using the transformation xᵢ = (1 − zᵢ)/2, where zᵢ ∈ {−1,+1}. The resulting coefficients describe interactions between quantum bits. In practice, software tools often perform this conversion, but professionals must still understand the modelling assumptions because an incorrect constraint encoding produces a technically valid answer to the wrong problem.

3. Quantum methods relevant to industry

Quantum annealing

Quantum annealing seeks low-energy states of an Ising or binary quadratic model. A physical or simulated annealing process gradually changes from an easily prepared initial system to a problem Hamiltonian. The measured low-energy states are candidate solutions. Quantum annealing is naturally aligned with QUBO-style models, although embedding a logical problem into hardware connectivity may require additional physical qubits.

The quantum approximate optimization algorithm

QAOA is a gate-based variational algorithm for combinatorial optimization. It alternates a problem-dependent operator with a mixing operator. A simplified state expression is:

Here, n is the number of qubits, p is the circuit depth, γ and β are adjustable parameters, U_C encodes the objective and U_B encourages exploration. A classical optimizer repeatedly changes the parameters using measurements from the quantum processor. The method is promising for constrained binary problems, but performance depends strongly on depth, noise, parameter optimization and problem structure.

Variational quantum algorithms and hybrid search

More generally, a variational quantum algorithm uses a parameterized circuit and a classical optimization loop. A realistic industrial architecture may combine a quantum sampler with classical repair, local search, tabu search, simulated annealing or mixed-integer programming. The quantum component may generate diverse candidate solutions while classical procedures restore feasibility and improve cost.

Quantum machine learning

Quantum machine learning may be investigated for classification, anomaly detection or forecasting. However, industrial data are often large, noisy, time-dependent and already well served by classical methods. A quantum model should be adopted only after establishing a clear hypothesis, a suitable data encoding, a fair benchmark and an operational benefit.

4. Logistics and vehicle routing

The vehicle routing problem (VRP) asks how a fleet should serve customers while minimizing distance, time or cost. Variants include the capacitated VRP, vehicle-routing-with-time-windows, pickup-and-delivery routing and electric-vehicle routing.

Let xᵢⱼᵏ = 1 when vehicle k travels from node i to node j. A distance objective is:

where dᵢⱼ is the distance between nodes i and j, measured in kilometres. Typical constraints require each customer to be visited once, conserve flow into and out of each node, respect vehicle capacity and eliminate disconnected subtours. If customer demand is qᵢ and vehicle capacity is Qₖ, a capacity condition for vehicle k is:

where yᵢₖ = 1 if customer i is assigned to vehicle k. This equation assumes demand is indivisible and vehicle capacity is measured in the same units as demand.

Time windows introduce arrival times. If service at customer i must occur between eᵢ and lᵢ, then:

where tᵢ is arrival or service time, in hours or minutes. Travel-time constraints link successive locations. In real applications, travel times are uncertain, so a route minimizing expected distance may perform poorly during congestion. Robust or stochastic optimization, scenario testing and real-time re-optimization are therefore essential.

Quantum approaches can encode assignment and sequencing decisions as QUBOs, but the number of binary variables can grow rapidly. A practical workflow may decompose a large route plan into geographic clusters, use a quantum method to improve a local subproblem, and then apply a classical route-feasibility and improvement procedure.

For electric fleets, routing must also consider charging. The objective may include distance, electricity price, battery degradation and delivery lateness. A route is feasible only if the state of charge remains within its limits and charging stations are available at appropriate times.

5. Workforce scheduling

Workforce scheduling assigns people to shifts while satisfying coverage, labour rules, qualifications, preferences and rest requirements. Let xᵢₛ = 1 if employee i works shift s. A basic coverage constraint for shift s is:

where rₛ is the required number of employees. If employee i is unavailable for shift s, the corresponding variable is fixed to zero. A cost objective might be:

where wᵢₛ is wage or assignment cost, P₁ and P₂ are penalties for undesirable violations, and λ₁, λ₂ are penalty weights. Hard legal requirements should normally be modelled as hard constraints or given very high protection; they should not be casually traded against small cost savings.

Quantum methods may help explore combinations of assignments, particularly when preferences and multiple skill categories create a dense search space. Yet a schedule is not acceptable merely because it has a low mathematical cost. Managers must check fairness, fatigue, accessibility, collective agreements and employee consultation. An automated schedule can reproduce historical bias if its objective rewards availability patterns produced by unequal working conditions.

6. Production planning and manufacturing

Production planning determines which products to make, in what quantities, on which machines and at what times. A simple capacity-constrained model is:

Here, mₚ is contribution margin per unit of product p, xₚ is production quantity, aₚ is machine time per unit and H is available machine time. Units must be consistent: if aₚ is hours per unit, H must be hours.

Real factories extend this model with setup times, material availability, batch sizes, due dates, sequence-dependent changeovers, work-in-process limits and maintenance windows. Binary variables may represent whether a machine is assigned to a job or whether a product changeover occurs. These interactions make the problem suitable for mixed-integer and potentially quantum optimization experiments.

In job-shop scheduling, each operation must be assigned to a machine and ordered relative to other operations. Quantum formulations may use binary variables for time slots or pairwise precedence decisions. Time-indexed formulations are conceptually straightforward but can require many variables. Alternative decompositions can reduce the model: for example, classical planning may determine production quantities, while a quantum or heuristic subroutine improves machine sequencing.

Manufacturing quality also creates a role for anomaly detection. Sensor data from vibration, temperature, acoustic emissions and power consumption can indicate process drift. Quantum machine learning is not automatically superior, but quantum-enhanced feature maps or kernels may be evaluated for carefully selected, moderate-sized classification tasks.

7. Energy systems and grid optimization

Power systems must maintain supply–demand balance while respecting network, generation, storage and reliability constraints. At a simplified level:

where Pᵍₜ is generation from generator g at time t, Pᵗ_dis is storage discharge, Pᵗ_ch is storage charging and Dₜ is demand. All terms must use the same power unit, such as megawatts. This balance alone is insufficient: generation limits, ramp rates, reserve requirements, transmission limits and state of charge must also be enforced.

Unit commitment decides which generators are on or off and how much they produce. Binary commitment variables create a difficult mixed-integer problem. A simplified generation-cost objective is:

where uᵍₜ ∈ {0,1} indicates whether generator g operates, Fᵍ is a fixed operating cost and cᵍ is variable cost per unit of energy. Startup costs and minimum up/down times make the model more realistic.

Quantum optimization may be applied to unit commitment, battery scheduling, microgrid operation, demand response and network restoration. A microgrid controller may choose when to charge batteries, use local generation or import electricity. If electricity price varies by time, the cost of purchased energy is:

where pₜ is price per kilowatt-hour and eₜ is imported energy in kilowatt-hours. The product has units of currency.

Forecast error is central. Renewable generation and demand are uncertain, so an apparently optimal schedule may become infeasible. Scenario-based optimization, reserve margins and receding-horizon control are safer than treating forecasts as exact. Grid applications also demand strong validation because an infeasible or unstable control decision can threaten public safety and system reliability.

8. Predictive maintenance

Predictive maintenance uses condition data to estimate failure risk or remaining useful life. A common reliability quantity is the survival function:

where T is the random failure time and S(t) is the probability that the asset survives beyond time t. The failure or hazard rate can be expressed as:

where f(t) is the probability density of failure time. These relationships apply when a suitable statistical model exists and the definitions of failure and observation time are clear.

Quantum computing may contribute in three ways. First, a quantum or hybrid classifier may identify abnormal sensor patterns. Second, optimization may select inspection dates, spare parts and technician assignments. Third, a quantum-inspired or quantum-assisted model may support portfolio decisions across many assets.

The business objective is not merely high predictive accuracy. A useful maintenance system balances false alarms, missed failures, inspection cost, downtime and safety. If a warning is issued too frequently, operators may ignore it; if it is issued too late, the system has little value. Maintenance decisions should therefore be evaluated using avoided downtime, maintenance cost, production availability and safety outcomes.

9. Industrial digital transformation

Quantum computing is one component of a broader industrial digital architecture. A mature system may include sensors and industrial control systems, an edge layer for low-latency processing, cloud or high-performance computing for planning, enterprise resource planning, a digital twin and a decision-support interface.

A digital twin is a continuously updated computational representation of an asset or process. Quantum optimization can be called by the twin to evaluate alternative schedules, routes or energy policies. The twin supplies current conditions; the optimizer proposes actions; classical validation checks constraints; operators approve or reject the recommendation; and outcomes return to the data system.

This arrangement illustrates an important principle: quantum computing should be integrated into a decision process, not deployed as an isolated demonstration. Interoperability, data lineage, identity management, auditability and human override are as important as circuit design.

A useful evaluation framework has four levels:

Quantum advantage should be defined carefully. A speedup claim must specify the problem class, input size, hardware, preprocessing, post-processing, baseline algorithm and quality of solution. Comparing a quantum method with a weak classical baseline is not evidence of meaningful advantage.

10. Illustrative case: a hybrid logistics and energy pilot

Background

Consider an illustrative regional food distributor operating 18 electric delivery vehicles from one depot. It serves 120 customers, uses time windows, and owns a solar-supported charging facility. Vehicle routes and charging schedules are planned separately, causing congestion at chargers and occasional late deliveries.

Problem

The company seeks to reduce total operating cost while maintaining delivery windows, vehicle capacity, battery limits and charger capacity. The decision variables include customer-to-vehicle assignments, route arcs, charging periods and imported electricity.

Analysis

The company first creates a classical baseline using a mixed-integer solver and a local-search route optimizer. Data quality is examined: customer demand units, travel-time estimates, charger availability and battery measurements are reconciled. The combined model is too large for a direct experimental quantum formulation, so the planning process is decomposed.

The classical system assigns geographic clusters and estimates feasible charging windows. A QUBO subproblem then selects among alternative route fragments and charging choices. A hybrid routine samples candidate solutions, repairs any violations, and applies 2-opt route improvement and a classical scheduling procedure. Here, 2-opt means exchanging two route edges when the resulting route is shorter and remains feasible.

Outcome

In a pilot, the hybrid method produces solutions within the same cost range as the best classical baseline, but it generates a more diverse set of alternatives under changing electricity prices. The company does not claim quantum advantage. The operational value comes from rapid scenario exploration and a better integrated view of routing and charging. Before deployment, the firm requires stress tests for congestion, vehicle failure, missing data and extreme weather.

Learning points

1. The business problem should be defined before selecting a quantum algorithm.
2. Decomposition is often necessary because the full industrial model is too large.
3. Classical repair and validation are essential.
4. A useful pilot measures operational outcomes, not merely circuit depth or sampling counts.
5. Safety and service commitments must remain hard constraints.

11. Challenges, limitations and responsible implementation

Hardware and algorithmic limitations

Current quantum devices are noisy, have limited qubit counts and may require costly connectivity transformations. Deep circuits can lose useful signal through decoherence and gate errors. Sampling variability means that repeated executions may produce different candidate solutions. Quantum annealers and gate-based devices also have different programming models, making results difficult to compare without careful experimental design.

Modelling limitations

Penalty formulations may generate infeasible solutions or distort the intended objective. Large industrial models may require variable reduction that removes important operational detail. Forecast uncertainty, incomplete sensor data and changing regulations can dominate any computational improvement.

Economic limitations

Costs include hardware access, software development, cloud execution, data engineering, integration, testing and specialist training. A small improvement in route distance may not justify a complex deployment. Total cost of ownership and opportunity cost must be included in the business case.

Safety and regulation

Energy and manufacturing decisions can affect public safety, workers and the environment. Critical systems require validation, fail-safe controls, access restrictions, incident response and human accountability. In many settings, an optimizer should recommend an action while an authorized operator or certified control system decides whether it is executed.

Fairness and workforce concerns

Scheduling systems can distribute undesirable shifts unfairly. Performance objectives should include transparent fairness constraints, such as limits on consecutive night shifts or equitable distribution of weekends, while respecting employment law and collective agreements. Workers should know what data are used and how decisions can be challenged.

Cybersecurity and intellectual property

Industrial control systems are attractive targets for attack. Quantum cloud access introduces additional credentials, interfaces and data-transfer pathways. Sensitive route, production and energy data may reveal commercial strategy. Security design should include least-privilege access, encryption, logging, supplier assessment and a plan for service interruption.

12. Best-practice implementation roadmap

A responsible organization can proceed through the following stages:

1. Select a valuable problem. Choose a problem with measurable cost, delay, energy or reliability impact.
2. Build a trustworthy baseline. Compare against leading classical solvers, not simple heuristics alone.
3. Audit the data. Check missing values, units, timestamps, outliers and ownership.
4. Formulate constraints explicitly. Separate hard safety or legal constraints from soft preferences.
5. Prototype classically and in simulation. Test the encoding before using quantum hardware.
6. Use hybrid decomposition. Assign quantum methods to subproblems where they are technically justified.
7. Validate feasibility and robustness. Test disruptions, uncertainty and worst-case scenarios.
8. Measure business outcomes. Include quality, runtime, cost, emissions, service and user acceptance.
9. Govern deployment. Establish audit trails, human oversight, cybersecurity and escalation procedures.
10. Scale only with evidence. A pilot should earn expansion through repeatable improvement.

Contemporary developments

Research is moving toward error mitigation, improved qubit connectivity, better compilation, quantum-classical co-design and problem-specific ansätze. Quantum-inspired optimization, tensor methods and advanced classical sampling are also producing useful techniques without requiring a fault-tolerant quantum computer. Industrial organizations should monitor these developments without assuming that every new algorithm has immediate commercial value.

The most realistic near-term pattern is coexistence. Classical optimization remains the operational foundation, while quantum processors may become specialized accelerators or sources of candidate solutions. Fault-tolerant quantum computers could eventually enable deeper algorithms and more reliable computation, but the timing and magnitude of industrial advantage remain uncertain.

Chapter Summary

Operations, logistics, energy and manufacturing involve complex decisions under constraints. Vehicle routing combines assignments, sequences, capacities and time windows. Workforce scheduling must balance coverage, qualifications, cost, fairness and labour rules. Production planning coordinates materials, machines, changeovers and due dates. Energy optimization must maintain power balance while handling generation limits, storage, transmission and uncertainty. Predictive maintenance combines reliability analysis, sensor data and intervention decisions.

Quantum annealing, QAOA and variational methods can represent or explore some of these problems, usually through binary quadratic models. Penalty terms can encode constraints, but their values and consequences require careful design. Hybrid quantum–classical workflows are currently the most practical approach: classical systems prepare data and enforce feasibility, while quantum methods may explore selected subproblems.

Successful industrial transformation depends on more than quantum hardware. It requires trustworthy data, strong classical baselines, robust validation, cybersecurity, human oversight, fairness and measurable operational benefit. Quantum advantage should be claimed only through transparent comparisons that include preprocessing, post-processing, solution quality and total cost.

Key Terms

Binary decision variable: A variable restricted to 0 or 1.

Constraint: A rule limiting the values or relationships of decision variables.

Digital twin: A computational representation that is updated using data from a physical asset or process.

Feasibility: The condition in which all required constraints are satisfied.

Hazard rate: The instantaneous conditional rate of failure given survival up to a specified time.

Hybrid quantum–classical algorithm: A method combining quantum computation with classical optimization or validation.

Ising model: A model using variables commonly represented as −1 and +1 with interaction coefficients.

Job-shop scheduling: Scheduling operations across machines subject to technological and capacity constraints.

Mixed-integer optimization: Optimization involving continuous, integer and/or binary variables.

NP-hard: A computational classification indicating that no generally efficient exact algorithm is known for all problem instances.

Objective function: A mathematical measure to minimize or maximize.

Penalty term: An added objective component that discourages constraint violations.

Predictive maintenance: Maintenance planning based on predicted asset condition or failure risk.

QAOA: Quantum approximate optimization algorithm, a parameterized gate-based optimization method.

QUBO: Quadratic unconstrained binary optimization model.

Quantum annealing: A quantum optimization approach seeking low-energy states of a problem Hamiltonian.

State of charge: The available energy in a battery relative to its usable capacity.

Subtour: A disconnected cycle in a routing solution that fails to form the required complete tour.

Unit commitment: Scheduling which power generators operate and at what output levels.

Review Questions

Short Answer Questions

1. What are the three essential components of an industrial optimization problem?
2. Why are penalty terms used in QUBO formulations?
3. State two constraints commonly found in vehicle routing.
4. What is the purpose of a digital twin in industrial decision support?
5. Why should quantum methods be compared with strong classical baselines?

Long Answer Questions

1. Explain how a vehicle-routing problem can be transformed into a binary optimization model. Include the role of assignments, route arcs and feasibility constraints.
2. Compare quantum annealing and QAOA in terms of representation, execution model and practical industrial considerations.
3. Discuss the technical and organizational requirements for quantum-assisted workforce scheduling.
4. Explain how uncertainty affects grid optimization and predictive maintenance decisions.
5. Evaluate the best-practice roadmap for introducing quantum optimization into a manufacturing organization.

Analytical / Application Questions

1. A distributor must serve customers with vehicle-capacity and delivery-window restrictions. Design a hybrid classical–quantum workflow and identify three metrics for evaluating it.
2. A factory’s proposed quantum scheduler reduces changeover time but occasionally violates maintenance windows. Should it be deployed? Justify your answer using safety, feasibility and governance principles.
3. A microgrid has solar generation, a battery and time-varying electricity prices. Propose decision variables, an objective and at least four constraints for a daily scheduling model.

Multiple Choice Questions with Answers

Question 1

Which expression best represents a linear routing cost when xᵢⱼ indicates whether travel occurs from i to j?

A. C = ∑ᵢ ∑ⱼ cᵢⱼ xᵢⱼ

B. C = ∑ᵢ cᵢⱼ + xᵢⱼ

C. C = ∏ᵢ ∏ⱼ cᵢⱼ xᵢⱼ

D. C = ∑ᵢ ∑ⱼ xᵢⱼ² − cᵢⱼ

Correct Answer: A. C = ∑ᵢ ∑ⱼ cᵢⱼ xᵢⱼ

Question 2

What is the main purpose of a penalty term in a QUBO formulation?

A. To remove all binary variables

B. To discourage constraint violations

C. To guarantee zero hardware noise

D. To convert costs into physical distances

Correct Answer: B. To discourage constraint violations

Question 3

In workforce scheduling, what does ∑ᵢ xᵢₛ ≥ rₛ represent?

A. A limit on total wages

B. A vehicle-capacity constraint

C. A minimum staffing requirement for shift s

D. A battery state-of-charge equation

Correct Answer: C. A minimum staffing requirement for shift s

Question 4

Which quantity must be balanced in the simplified power-system equation?

A. Generation, storage actions and demand

B. Employee skills and wages

C. Vehicle distance and customer number only

D. Machine temperature and vibration only

Correct Answer: A. Generation, storage actions and demand

Question 5

Which statement most accurately describes QAOA?

A. A purely classical sorting algorithm

B. A gate-based variational algorithm alternating problem and mixing operators

C. A database protocol for industrial sensors

D. A method that requires no classical optimization

Correct Answer: B. A gate-based variational algorithm alternating problem and mixing operators

Question 6

If a vehicle has capacity Qₖ and assigned demand totals 900 kg, which condition is necessary for feasibility?

A. Qₖ must be less than 900 kg

B. Qₖ must equal zero

C. Qₖ must be at least 900 kg

D. Qₖ must be measured in hours

Correct Answer: C. Qₖ must be at least 900 kg

Question 7

What is the survival function S(t) in reliability analysis?

A. The probability that failure occurs exactly at t

B. The probability that the asset survives beyond t

C. The cost of replacing an asset at t

D. The number of sensors installed at t

Correct Answer: B. The probability that the asset survives beyond t

Question 8

Which is a best practice when evaluating a quantum industrial pilot?

A. Compare it only with an intentionally weak classical method

B. Ignore preprocessing and post-processing time

C. Measure operational outcomes against a strong classical baseline

D. Treat every sampled solution as automatically feasible

Correct Answer: C. Measure operational outcomes against a strong classical baseline

Question 9

Why can excessive penalty coefficients be problematic?

A. They can distort numerical balance in the objective

B. They eliminate all operational constraints

C. They guarantee a classical speedup

D. They make all solutions equally optimal

Correct Answer: A. They can distort numerical balance in the objective

Question 10

Which issue is especially important when automating employee schedules?

A. Fairness and compliance with labour rules

B. Eliminating all human review

C. Using distance in kilometres for wages

D. Ignoring employee availability

Correct Answer: A. Fairness and compliance with labour rules