Optimise

Use the model to make the machine better

Once the model can be trusted, it does the searching. It tries thousands of designs against the number that matters, whether that is lap time, mass, cost or throughput, and every result is checked in the full model before anyone believes it.

Why every result is checked

A fast model finds any gap in itself at once. On the GT car the quick lap model promised 3.1%; driven in the full simulation the gain was 1.2%. That gap is why nothing is trusted until the full model has driven it.

Optimised so far

  1. Vehicle dynamics

    A GT car, redesigned for lap time

    It locked the differential, added a stiff rear third spring, raised both roll centres and took wing off for less drag. 1.2% off the best driven lap.

    See it
  2. Urban Aerial Defence

    Six airframes, a hundred attacks

    Interceptor airframes modelled from first principles at equal mass and power, and flown against the same attacks over a city.

    See it
  3. Robotics

    A quadruped from a task description

    Body, controller and structure designed together: 288 bodies searched in about 2.5 hours on one GPU, ending in printable geometry.

    See it

Have a problem like these?