Extending Human–Machine Interaction Analysis from Autonomous Driving to Manned–Unmanned Vehicle Teaming: A Function-Specific Effectiveness Framework and the Partial-Autonomy Trap

Human–machine interaction (HMI) has become a central issue in autonomous driving because partial automation can degrade, rather than improve, human performance during takeover, handover, and out-of-the-loop transitions. Similar interaction risks are emerging in manned–unmanned vehicle teaming (MUM-T), where operators must supervise multiple autonomous or remotely controlled assets under higher mission complexity and safety-critical constraints. However, existing effectiveness analyses of unmanned and MUM-T systems often treat the level of autonomy (LOA) as a fixed system attribute or assume the highest autonomy level, thereby obscuring the human–automation bottlenecks that arise during partial autonomy. This study proposes a function-specific HMI effectiveness framework in which autonomy is represented as a vector across surveillance, maneuver, fire or neutralization, command and control, and human–machine teaming functions. Measures of performance (MOPs) are modeled as conditional performances jointly shaped by function-specific LOA, operational environment, and intrinsic system capability, and are propagated through a five-layer LOA–MOP–MOE structure to a mission-level measure of effectiveness (MOE). The framework is demonstrated using a notional mine countermeasure scenario in which manned minehunters cooperate with unmanned underwater and surface vehicles. Four autonomy progression stages, from manned-centric operation to advanced cooperative autonomy, are evaluated for timely route opening. The case illustrates a non-monotonic partial-autonomy trap, or LOA-2 valley: remotely controlled unmanned assets may temporarily reduce mission effectiveness when teleoperation workload and HMI bottlenecks outweigh equipment gains, before cooperative and supervisory autonomy restore and exceed baseline performance. The contribution of this study lies not in the notional numerical results but in providing an explicit diagnostic structure for identifying where and why function-specific autonomy, control sharing, and HMI bottlenecks shape mission effectiveness. The framework thereby extends human–machine interaction analysis from autonomous driving to the broader, higher-risk setting of manned–unmanned vehicle teaming.