Shared, persistent site record for people and robots

A diagnosis model for cherry and zelkova already existed, reading bark condition, vigour and fungal bodies from four part views per tree. The question was the one upstream of it: how should tens of thousands of trees actually be photographed, by crews carrying phones, helmet cameras, chest cameras or nothing at all? That looks like logistics and is optics. It is decided by how many pixels land on a millimetre of bark, and whether they moved during the exposure.

A handling robot has to know a bag's three edges, its rigidity and its surface material before the bag reaches it, and none of it is in the airline's systems, because the tag carries a booking rather than a shape. The competition line runs at 25 m/min and is never stopped, around eighty pieces go in continuously and often at an angle, and the rows we write are matched to the truth by insertion order alone. There is no identifier anywhere in the scene, which means the count is the identity: lose it once and every row after it is attached to the wrong bag, however well it was measured.

A lithium-ion cell arrives at a waste site inside something else and is almost never declared. It survives the household bin and not the compactor, the grapple or the shredder; when it fails it fails in seconds, needs no outside oxygen, and the material around it is combustible and deep. The operator came with the framing the market offers: buy a detector.