Open AccessTutorial on the generation of ergodic trajectories with projection‐based gradient descent
Author(s) -
Dressel Louis,
Kochenderfer Mykel J.
Publication year - 2019
Publication title -
iet cyber‐physical systems: theory and applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.308
H-Index - 7
ISSN - 2398-3396
DOI - 10.1049/iet-cps.2018.5032
Subject(s) - ergodic theory , ergodicity , trajectory , projection (relational algebra) , computer science , gradient descent , projection method , mathematics , control theory (sociology) , mathematical optimization , algorithm , control (management) , mathematical analysis , artificial intelligence , dykstra's projection algorithm , physics , statistics , astronomy , artificial neural network
A vehicle trajectory is ergodic with respect to some spatial distribution if the time spent in a region is proportional to the region's integrated density. One method of generating ergodic trajectories is projection‐based trajectory optimisation, a gradient descent method that supports non‐linear dynamics and balances trajectory ergodicity with control effort. In this study, the authors survey the existing literature on projection‐based trajectory generation, focusing on implementation. They introduce an easy‐to‐use, open‐source software package that generates ergodic trajectories orders of magnitude faster than previously reported. The authors’ goal is to simplify the implementation of projection‐based ergodic control so it can be applied widely across disciplines.