The headlines about a potential new MH370 seabed search talk about hope and high-tech robots. As a pilot who has spent long hours thinking about what happens when an aircraft disappears from the radar picture, I look at two things first: the quality of the evidence that refines where to look, and the practical capabilities of the tools that will be sent into the deep. Neither alone finds wreckage. Together, they change the odds.
In May 2024 Ocean Infinity presented a formal proposal to Malaysian authorities to resume an underwater search based on updated analysis. That submission is important because it confirms the company believes it has new inputs worth testing, and because Ocean Infinity is proposing to operate on a no-find, no-fee basis familiar from its 2018 effort.
What has actually changed on the equipment side since 2018 is not magic. It is incremental but meaningful: longer-range, higher-resolution autonomous underwater vehicles, wider swaths of synthetic aperture sonar coverage, better endurance from long-endurance AUV platforms, and more mature automated processing to help triage the millions of sonar images an AUV generates. Those are the building blocks that turn a broad search box into something you can realistically sweep to a useful confidence level.
A practical baseline: modern HUGIN-class AUVs have been developed with endurance and payloads suited for deep, rugged Atlantic and Indian Ocean missions. Kongsberg and others have pushed long-range models through factory and sea trials in 2023 and early 2024, emphasizing dual-use sonar suites that carry high-resolution synthetic aperture sonar, multibeam echosounders, and sub-bottom profilers. In operational terms that means a single vehicle can collect imagery at resolution and over ranges that used to require far more time and a much larger platform footprint. For search managers that improves coverage per unit time and reduces the number of launches and recoveries, which are the riskiest parts of any AUV operation.
Ocean Infinity’s corporate model has also shifted toward a standardized Armada vessel design to host and support larger AUV fleets. These purpose-built support ships, delivered to the market beginning in 2023, are optimized for autonomous operations, which helps keep more vehicles in the water longer and reduces weather downtime when sea states permit safe launch and recovery. That matters in the Southern Indian Ocean where weather windows are constrained. The hardware alone does not overcome seasonal rough seas, but better platform design and mission planning can lengthen effective survey time.
On the data side, two trends matter: sensor fusion and automated triage. Search AUVs now commonly carry complementary sensors that cross-validate a contact: SAS or sidescan to image a contact, multibeam to map its shape and shadow, and magnetometers or sub-bottom profilers to detect buried metallic objects. Where an object shows up on multiple sensors, you escalate. Where only one sensor shows a faint anomaly, you flag it for secondary inspection or discard it as background noise. That tiered approach conserves the most expensive resource: manned time using an ROV for visual identification. The same approach underpinned previous large searches but the sensor packages and procedural discipline are tighter now.
Machine learning and modern signal processing are not replacing human analysts, but they are reducing the bottleneck. Automated target recognition applied to sidescan and SAS imagery has matured in academic and applied work over the last few years; convolutional neural networks and ensemble methods can raise detection rates and reduce false positives when trained on varied datasets. In a practical search this matters because a fleet of AUVs can generate terabytes of imagery in weeks. Any manual review pipeline becomes untenable without automated pre-sorting. That said, these algorithms have limits. They are trained on particular target shapes and seabed textures, and their real-world performance drops when the seafloor is rough, when targets are heavily broken, or when debris is partially buried or covered in biota. Expect human analysts to still be the final arbiters for any candidate wreckage.
There is one more data angle people keep bringing up: passive radio-signal techniques such as WSPR. Some researchers propose combining those results with Bayesian particle-filter approaches used in earlier MH370 work to nudge search probability maps. That is an interesting academic avenue and it may provide additional priors to weight portions of a search box, but WSPR-based inferences are still novel and controversial. Any operational search will need to treat those inputs as probabilistic cues rather than deterministic fixes. In plain terms: new signals can point you to places you did not previously prioritize, but they will not substitute for acoustic imaging when you get on site.
What this means for the families and for aviation is straightforward realism. A narrower, better-justified search area combined with improved AUV endurance, higher-resolution synthetic aperture sonar, and smarter automated processing improves the probability of detection compared with the same search a decade ago. It does not guarantee success. The southern seafloor is extremely complex, and wreckage can disintegrate, be buried by sediment, or be scattered by currents. And real-world factors such as fishing vessel activity and seasonal weather windows impose operational ceilings. The 2018 Ocean Infinity mission demonstrated both the promise and the limits of the approach. Lessons from that effort appear to be reflected in the more targeted proposal presented in 2024.
From an operational standpoint my checklist for any credible renewed effort would include these items: a transparent statement of the evidence that narrowed the search box, a clear sensor suite and survey plan showing how AUV endurance and swath width translate to daily area coverage, a post-survey triage workflow describing how automated detections are validated, and an honest assessment of seasonal constraints and contingency plans. Those are the details that separate a public relations exercise from a real chance at closure.
If the search goes ahead, the aviation community should watch two things closely: how the priors for the search area are justified, and how the survey handles the inevitable long tail of ambiguous contacts. Those are the technical and operational issues that determine whether we are looking harder, or simply looking again. The equipment has improved. The hard work remains in turning better sensors and smarter analytics into reliable, verifiable identifications on the seafloor.