Incorporating the Human Element in UAV Tasking

An operator contols a remotely piloted aircraft. (

Despite lacking an on-board pilot, UAVs require crews of up to three human operators. These crews are already experiencing high workload levels, a problem that will likely be compounded as the military envisions a future where a single operator controls multiple UAVs. To accomplish this goal, effective scheduling of UAVs and human operators is crucial to future mission success. We present a mathematical model for simultaneously routing UAVs and scheduling human operators, subject to operator workload considerations. This model is thought to be the first of its kind. Numerical examples demonstrate the dangers of ignoring the human element in UAV routing and scheduling.

In this work, human operators are required to perform three activities associated with UAV missions: Pilot the aircraft while over a target, pilot an aircraft while flying between targets, and analyze data captured from target locations. The problem is to simultaneously maximize the overall effectiveness of the mission while balancing human operator workloads. Results from this work will be leveraged in the establishment of appropriate new metrics for dynamic communications networks.

The mathematical representations of operators’ multi-task workloads and workload capacity limits are predicated upon existing mental workload theories and methods. In particular, the resource theories propose that the human has a limited capacity (resources) for processing information and this capacity can be allocated in graded amounts to different activities. These theories view performance decrement as a shortage of resources; if an operator simultaneously performs multiple tasks and their combined demands exceed the operator’s capacity, the performance of some or all of the tasks may suffer. The visual, auditory, cognitive and psychomotor (VACP) method provides a computational framework for aggregating individual workloads of multiple simultaneous activities to represent the combined workload. The method assumes that individual workloads of multiple simultaneous activities at a time instant can be linearly added together to represent the instantaneous combined workload. Each activity’s workload is pre-estimated based on subject matter experts’ subjective ratings.

Operator workload may increase in a nonlinear manner as additional tasks are performed.

Figure 1: Operator workload may increase in a nonlinear manner as additional tasks are performed.

In addition to the linear sum of individual workloads of simultaneously performed activities, the representation of the total mental workload in this paper also considers the cost of concurrence component, which represents the additional workload purely due to having to manage or supervise multiple activities. The cost of concurrence is expressed using a scaling parameter, Rn; it represents the additional workload due to “supervising” or “switching” activities when an operator performs n activities simultaneously. The cost of concurrence may increase in a linear or nonlinear manner as operators are burdened with more tasks. One representative functional form for this additional workload is depicted in Figure 1.

Each human operator, h, is assumed to possess a maximum workload threshold, represented by Whmax, which corresponds to the mental capacity concept of the resources theories.  At any time instant, each operator’s multitask workloads must fall below this threshold.

Figure 2: Deviations from the target workload.

Figure 2: Deviations from the target workload.

While excessive workloads can result in decreased operator performance, it has also been observed in the literature that operators perform best with some degree of stimulation.  Therefore, we incorporate the notion of a target workload level, as depicted in Figure 2.  Deviation from this target level will result in a penalty in the overall mission effectiveness.

The following small-scale example highlights the importance of balancing human-machine workload in the context of UAV mission planning. An intelligence-gathering mission involving five targets is considered, where targets 1, 3, and 5 are of higher priority. Two identical UAVs are available to observe these time-sensitive targets. A single operator is responsible for two tasks: Piloting the UAV as it observes a target, and analyzing data captured from the UAV to classify each target.  The piloting (control) activity must be performed simultaneously with each UAV; the analysis activity may be performed later (e.g., watching recorded video).  The operator may experience a non-linear increase in stress as additional activities are performed concurrently.  Figure 3 shows the optimal assignments of UAVs to targets when operator workload concerns are ignored.  Note that at any time the operator may be assigned to perform as many as six activities simultaneously. If the operator is unable to manage these tasks concurrently, this mission plan is not viable.



Figure 3: A UAV mission plan, with corresponding operator workloads, where operator capacities are ignored.















Conversely, Figure 4 shows an alternate plan, where the operator’s workload threshold has been explicitly considered.  Given this threshold, the operator is unable to perform more than one activity at a time.  Of note, only one UAV is employed, only three of five targets are observed, and those targets are observed later within their time windows.  This small example of human-machine optimization indicates that incorporating operator workload can lead to significantly different UAV assignments.





Figure 4: Optimal UAV and operator assignments, where operator workload thresholds are observed.