Operator Workload Modeling in Multi-Vehicle Teleoperation: A Simulation-Based Study

Abstract

The proliferation of unmanned systems has created scenarios where single operators must supervise multiple vehicles simultaneously, raising critical questions about cognitive workload limits, performance degradation, and optimal task allocation strategies. This paper presents a comprehensive simulation-based study of operator workload in multi-vehicle teleoperation, developing predictive models that inform system design and operational procedures. We systematically examine workload components in multi-vehicle supervision including perceptual demands from monitoring multiple video feeds and sensor displays, cognitive demands from maintaining situation awareness across vehicles and planning coordinated actions, and motor demands from issuing control commands and managing interface interactions. The study develops a multi-fidelity simulation framework integrating vehicle dynamics models (fixed-wing UAVs, quadrotors, ground robots), mission scenarios (surveillance, search and rescue, convoy operations), and human operator models based on established cognitive architectures (ACT-R, IMPRINT). We implement multiple workload assessment methodologies including task analysis-based prediction using techniques like NASA-TLX adapted for simulation, timeline analysis identifying periods of concurrent task demands, and computational cognitive modeling that simulates operator attention allocation, working memory limitations, and decision-making processes. The simulation environment systematically varies factors hypothesized to influence workload: number of vehicles under supervision (1-10), vehicle heterogeneity (homogeneous swarms versus mixed platforms), autonomy level (manual control, waypoint navigation, supervised autonomy), mission complexity (independent tasks versus coordinated maneuvers), and interface design (separate displays versus integrated common operating picture). Performance metrics encompass mission success rates, response time to critical events, vehicle utilization efficiency, and operator errors including missed alerts, incorrect commands, and mode confusion. Simulation results reveal a nonlinear relationship between vehicle count and workload, with manageable workload up to 3-4 vehicles under manual control but graceful scaling to 8-10 vehicles when high-level supervisory control is available. We identify critical workload bottlenecks including attention switching costs when alternating between vehicles (approximately 2-3 seconds per switch), working memory limitations in tracking vehicle states and mission progress, and decision-making delays when prioritizing between competing demands. The study examines workload mitigation strategies through simulation experiments: adaptive automation that increases autonomous capability when operator workload is high, intelligent alerting systems that filter and prioritize notifications based on urgency and operator availability, and task allocation algorithms that distribute responsibilities to balance workload. Comparative analysis demonstrates that adaptive interfaces providing workload-appropriate information density reduce peak workload by 30-40% compared to static displays. We investigate individual differences in multi-vehicle supervision capability, simulating operator populations with varying expertise levels, spatial ability, and multitasking capacity, revealing that expert operators manage approximately 50% more vehicles than novices at equivalent performance levels. The paper examines training implications, using simulation to evaluate training protocols including part-task training focusing on individual skills versus whole-task training with complete multi-vehicle scenarios. Advanced modeling techniques are employed including machine learning approaches that predict workload from physiological signals (heart rate variability, eye tracking) enabling real-time workload assessment, and reinforcement learning models of operator behavior that capture adaptive strategies for managing high workload. The study analyzes failure modes in multi-vehicle teleoperation including workload-induced tunneling where operators focus on one vehicle while neglecting others, and cascade failures where problems with one vehicle consume attention allowing issues to develop in others.