Generating 'big data' for manoeuvrability studies with trajectory optimization

While steady-state gait is relatively well-understood, transient manoeuvres have received far less dedicated research. These motions are difficult to study because they can be dangerous when executed at high speed. Additionally, differing points of initiation, animal morphologies and environments create a seemingly insurmountable variety of possible cases. Trajectory optimization of dynamic models is a possible means of overcoming these challenges, as it allows large sets of simulated manoeuvres to be generated under specified, repeatable conditions in a controlled environment. We describe a state-of-the-art trajectory optimization method and demonstrate that data generated in this way can be used to identify fundamental motion patterns for executing rapid manoeuvres in bipeds and quadrupeds under different conditions. Further, we compare the simulated manoeuvres to steady-state gait and discuss whether transients can be effectively modeled as modifications of steady-state motion patterns.