Finding a needle in a black ocean
The most famous shipwreck on Earth hid for 73 years. What finally found it was not a better submarine — it was a better search.
By 1985 the Titanic had defeated four decades of expeditions. She lay somewhere in a search box of hundreds of square kilometres, under 3,800 metres of water, in permanent darkness. Robert Ballard's team towed Argo — an unmanned camera sled — behind their ship in long parallel lines, a pattern search crews call "mowing the lawn", with crews watching grainy video through the night.
The insight that cracked it was about evidence, not optics. A sinking ship sheds a trail: coal, fittings, debris scattered by currents as it falls. A debris field is kilometres wide — a far bigger target than a hull. Ballard hunted the trail, not the ship. In the small hours of September 1, 1985, a boiler drifted through Argo's camera, and the trail led them to the bow.
No one aboard touched the wreck that night. The discovery belonged to a towed robot, a clever search pattern, and people who understood what the seafloor would remember about a sinking. That is the quiet lesson: underwater, finding is a harder problem than reaching.
Sense, plan, act — and how robots are wired
Every robot runs the same loop. Sense: sonar, cameras, pressure, compass. Plan: where am I, what am I looking for, where next. Act: thrusters, winches, arms. Argo was the sense; the plan lived in human heads on the ship; the act was the ship's own propellers dragging the sled. The loop does not care where each part runs.
Underwater, the senses split by range. Sonar sees far — kilometres — but fuzzily: a wreck is a bright smear on a scan. Cameras see truth but only metres of it. So real searches are layered: sonar to find candidates, cameras to confirm them. Wide and vague, then narrow and sure.
Modern robot software mirrors the loop in code. In ROS — the Robot Operating System, the framework we will build with in later weeks — each job is a small program called a node (sonar driver, detector, pilot), and nodes talk by publishing messages on named topics. Argo's whole expedition, redrawn as nodes and topics, fits on a napkin. Hold onto that picture; we will build it for real.
A sonar trace rolls past with targets buried in noise. Set the detection threshold yourself: too low and false alarms flood the pilot, too high and the wreck slips by. Watch each ping travel the node graph underneath.
Fly to the North Atlantic grid where Argo's cameras caught that first boiler in 1985 — the Titanic's resting place.