Scheduling ground handling operations with a constraint solver

An operational environment with zero margin for error

Ground handling is the set of services performed on an aircraft between its arrival at the gate and its departure for the next flight. This turnaround window is short, often between 30 and 90 minutes, and every minute counts. The work spans multiple functions: check-in, security screening, ramp operations, baggage handling, escort services, cargo processing, catering loading, document verification and gate management. Each function requires dedicated staff with specific training and certifications.

What makes ground handling scheduling particularly demanding is the variability. The flight programme changes by day of the week, by season, and sometimes with very short notice. A Monday may have 40 departures, a Tuesday 55. Charter operations appear and disappear. Airlines adjust frequencies. Each change in the flight schedule ripples directly into staffing requirements, because the number of agents needed at any given time depends on the number and type of aircraft being handled simultaneously.

Planners in ground handling organisations build monthly rosters for teams of 50 to 300 agents across these functions. They must account for qualifications, legal rest rules, contractual hours, absences, part-time contracts and operational coverage targets, all at once. Doing this manually with spreadsheets is possible for small teams, but it becomes increasingly fragile as headcount grows and flight programmes become more complex.

The qualifications matrix

In ground handling, not every agent can perform every task. A ramp agent is trained and certified for aircraft loading, pushback coordination and baggage sorting. A security agent holds an aviation security certification that allows them to screen passengers or cargo. A check-in agent is trained on airline-specific departure control systems. A document check agent is authorised to verify travel documents at the gate. These qualifications are regulated by aviation authorities and cannot be bypassed.

The qualification matrix defines, for each agent, the set of positions they are authorised to fill. Some agents hold a single qualification. Others are cross-qualified and can work across two or three functions. When the solver builds a monthly roster, it treats qualification matching as a hard constraint: an agent can only be assigned to a shift position if they hold the required qualification. There is no override, no exception, no soft fallback. If an agent is not qualified for a position, the solver will never assign them to it.

This is a fundamental difference from manual scheduling. A planner working under time pressure may occasionally assign an agent to a position they are not fully certified for, intending to fix it later. A constraint solver cannot make this mistake. The qualification matrix is encoded once and enforced on every assignment, every day, for every agent in the roster.

Daily staffing needs driven by the flight schedule

Ground handling staffing needs are not static. They are derived directly from the flight schedule. On a day with 12 wide-body departures, the ramp team needs more agents than on a day with 8 regional flights. When a new airline launches a route, the check-in and security needs increase. When a seasonal charter programme ends, they decrease.

This means that the staffing requirement is defined day by day, position by position, for the entire month. Monday may require 8 ramp agents, 4 security agents and 6 check-in agents. Tuesday may require 12, 5 and 8. The solver receives these daily needs as input and attempts to assign agents so that every position is covered on every day.

Coverage is modelled as a soft objective with a heavy penalty. If the solver cannot find an assignment that covers all positions, it will still produce a roster, but the uncovered position carries a significant penalty that pushes the solver to explore alternatives. The planner sees exactly which positions are short-staffed and can make informed decisions about overtime, temporary staff or operational adjustments.

This approach is far more transparent than manual scheduling, where coverage gaps are often discovered only when the roster is published or, worse, on the day of operations.

Legal rest periods and forbidden shift sequences

Labour regulations in ground handling require a minimum rest period between consecutive shifts, typically 11 hours. This rule is straightforward to state but surprisingly difficult to enforce manually across a full monthly roster. The problem is not individual shifts in isolation. It is the interaction between shifts on consecutive days.

Consider an agent who finishes a late evening shift at 23:00. They cannot start an early morning shift at 05:00 the next day, because the gap is only 6 hours. But they could start a shift at 10:00. The set of forbidden sequences depends on the exact start and end times of every shift in the system. For an operation with 30 or 40 distinct shift types, the number of incompatible pairs can run into the hundreds.

The solver pre-computes all incompatible shift pairs before the optimisation begins. For every combination of a shift on day d and a shift on day d+1, it checks whether the rest period constraint is satisfied. If not, the pair is forbidden as a hard constraint. The solver then guarantees that no agent is ever assigned a forbidden sequence, regardless of how complex the shift structure is.

This eliminates an entire category of compliance risk that is endemic to manual rostering. Rest period violations in manual schedules are rarely intentional. They arise from the difficulty of tracking hundreds of shift transitions across dozens of agents over a full month.

Night shifts, weekends and fairness

Night shifts and weekend duties are the most constraining assignments in ground handling. They affect work-life balance, are subject to additional regulatory requirements, and are often a source of friction between staff and management. A fair distribution of these assignments is both a legal obligation and an operational necessity for retention and morale.

The solver addresses this through several mechanisms. First, it guarantees at least one free weekend per month for every eligible agent. This is modelled as a soft constraint with a high penalty, meaning the solver will sacrifice other objectives to preserve it. Second, it limits consecutive working days to a maximum of six, preventing the fatigue accumulation that results from extended work stretches without rest.

For shift equity, the solver tracks how many shifts of each type are assigned to each qualified agent within a functional group. If three agents are all qualified for night shifts, the solver distributes night assignments as evenly as possible among them. The equity calculation excludes agents who are on holiday or vacation during part of the month, so that absent agents do not distort the distribution for those who are present.

Part-time and student contracts are handled differently. Students may be excluded from the weekend guarantee while still participating in the general shift pool. Auxiliary staff may have maximum hour limits that the solver respects as hard constraints. Each contract type carries its own set of rules, and the solver applies them individually.

Functional groups and cross-qualification

Ground handling teams are typically organised into functional groups: ramp, security, check-in, cargo, catering and others. Each group has a primary function and a pool of agents who specialise in that function. However, many agents hold qualifications that span multiple functions. A ramp agent may also be qualified for cargo. A check-in agent may hold a document verification certification.

Cross-qualification is operationally valuable because it provides flexibility. When one function is overstaffed relative to its needs and another is short, a cross-qualified agent can fill the gap. The solver leverages this automatically. It assigns agents to their primary group by default, but when coverage gaps appear in another function, it can reassign cross-qualified agents to fill them.

The balancing act is subtle. If the solver moves too many agents out of their primary group, workload equity within that group suffers. If it moves too few, coverage gaps persist. The solver handles this by applying equity penalties within each group alongside the coverage penalties. The result is an assignment that fills as many positions as possible while keeping the workload distribution fair within each team.

This trade-off is exactly the kind of multi-dimensional optimisation that manual planners struggle with. A planner can solve one coverage gap by moving an agent, but they cannot easily see the downstream effect on equity, rest periods and next-day assignments. The solver evaluates all of these simultaneously.

From flight schedule to monthly roster

The end-to-end workflow in a ground handling operation follows a clear sequence. The flight programme for the upcoming month is received from the airport or airlines. This programme defines which flights operate on which days, their scheduled times, and the aircraft types involved. From this programme, the planning team derives the daily staffing needs: how many agents of each function are required on each day.

These needs are entered into the system alongside the employee data: the list of agents, their qualifications, contract types, known absences, holidays and any fixed scheduling constraints. The solver then takes all of this as input and generates an optimised monthly roster in minutes.

The output is a complete assignment: every agent, every day, every shift. The planner reviews the roster, checks the coverage summary, examines any flagged shortages, and can make manual adjustments where operational judgment is needed. The solver does not replace the planner. It handles the computational complexity so the planner can focus on the decisions that require human expertise: managing exceptions, handling last-minute changes, and communicating with the team.

The entire process, from flight schedule to published roster, moves from days of manual work to a structured workflow that takes hours. The compliance checks are automatic. The equity distribution is consistent. The coverage is globally optimised rather than locally patched. And every assignment can be traced back to the constraints and objectives that produced it.