A Fleet With a Maintenance Problem
The United States Navy's surface fleet has been struggling with a maintenance backlog for years. Ships wait months longer than scheduled for yard repairs, readiness rates have declined across key surface combatant classes, and the problem has drawn sustained criticism from Congress and fleet commanders alike. The service has now taken a concrete step toward addressing the underlying diagnostic challenge: the Navy does not always know the extent of a ship's structural deterioration until it is already in the shipyard, at which point unexpected repairs cascade into extended stays and ballooning costs.
Gecko Robotics, a Pittsburgh-based company specializing in deploying robotic inspection systems on industrial infrastructure, has been contracted to help close that information gap. The five-year, $54 million indefinite delivery, indefinite quantity contract will deploy the company's AI-enabled robots across 18 ships assigned to the Navy's Pacific Fleet, with the goal of identifying maintenance needs earlier and more comprehensively than traditional manual inspection allows.
How Gecko's Technology Works
Gecko's robotic systems are magnetic-track crawlers capable of scaling vertical metal surfaces — including ship hull sections, bulkheads, and tank walls — while carrying an array of sensors. Ultrasonic thickness gauges measure steel plate thickness at thousands of points per hour, detecting corrosion and metal loss that would take human inspectors days to map manually. Thermal imaging sensors identify hot spots that may indicate bearing wear, insulation degradation, or electrical faults. High-resolution cameras document surface conditions with visual fidelity that supports both immediate decision-making and historical trending.
The robots feed collected data into an AI analysis platform that processes sensor streams in near real time and flags anomalies against baseline measurements from previous inspections. For ship systems, this means maintenance crews receive a ranked list of areas requiring attention, with severity estimates derived from the rate of deterioration rather than a single snapshot. Predictive maintenance — identifying that a component will fail within a defined window rather than waiting for it to fail — requires exactly this kind of longitudinal data collection.
