Experimental study on the impact of long communication delays on autonomous decision-making in deep space habitats

Space vehicles and habitats are complex systems with tightly coupled interdependencies. They operate under extreme, unforgiving conditions that challenge system performance and crew safety. Effective health management in such systems demands strong architectural and situational knowledge, leading to heavy reliance on ground control. As missions venture deeper into space, growing communication delays make such support impractical, prompting a paradigm shift toward greater on-board autonomy. However, computational constraints limit the migration of all health management components to the habitat. Running high-fidelity fault detection and diagnosis modules on the ground can conserve on-board resources but introduces delays in decision-making. This latency increases the need for effective decision-making policies, which may require interrupting agents to enable timely responses to higher-priority situations. In this paper, we study how communication delays and agent interruptibility impact decision-making in a habitat system, using the Human-centered Autonomous Resilient Space Habitat (HARSH), a reconfigurable cyber-physical testbed. A series of experiments is conducted to evaluate different design configurations under single and compound disruption scenarios. The ability of each configuration to respond and recover from faults is assessed using resilience metrics. Findings from the experiments inform recommendations to advance autonomous decision-making policies and enhance system resilience in future deep space habitats.