There were two main aspects to Manchester University’s study. The first was digitizing the dinosaur and creating a 3D simulation in the computer. This is something Sellers’ team has been working on for a number of years. Way back in 2013, Sellers used then-new laser-scanning techniques and an advanced computer modeling system to recreate the Argentinosaurus, a friendlier, herbivorous dinosaur that measured a whopping 131 feet and weighed as much as 80 tons.
The second part of the T. Rex study is the really groundbreaking bit, though. This is where the machine learning comes in, which was used to generate the necessary locomotion calculations and incorporate myriad hard and soft constraints in the simulation.
“[I utilized] completely unsupervised learning, using a goal criteria of ‘Get from A to B as fast as possible,’ and given enough computer time it can manage that all on its own,” Sellers explained. “This gets around the problem of just getting out what you put in. No one knows how T. Rex, moved but at least this way you have an objective optimization criteria to justify the choices made. It also means that it is relatively easy to add in extra constraints like ‘don’t break any bones.’ It’s basically a multi-physics, multi-objective simulation system.”
In the conclusions reached by the researchers, the T. Rex could not run due to its size and weight. This means it would have been unable to pursue prey at high speeds, and even its regular walking speed would have been hampered due to its unwieldy skeleton.
Click here to read the full article.