FAME-QP: A GPU-Accelerated Quadratic-Programming Solver for Real-Time Model Predictive Control

研究论文 Hugging Face Papers 原文语言:中文

论文摘要(作者报告,未经本站复测):We present FAME-QP, a GPU-accelerated solver for dense quadratic programs arising in linear model predictive control.

全文

WepresentFAME-QP,aGPU-acceleratedsolverfordensequadraticprogramsarisinginlinearmodelpredictivecontrol.FAME-QPcombinesconstraint-dependentoperatorrepresentationswithfusedsingle-andtwo-kerneliterationsreplayedthroughCUDAgraphs,automaticallyselectingtheexecutionconfigurationthroughdevicecalibrationandproblem-specifictiming.Itprobesacachedoperatorbanktoselecttheinitialpenaltyandupdatesoperatorslazilywhenproblemmatriceschange.NumericalbenchmarksandsimulatedhumanoidandquadrotorcontrolexperimentsdemonstratereducedlatencyrelativetocompetingGPUsolverswhilemaintainingsolutionaccuracyandcontrolperformance.

View arXiv pageView PDF Project page GitHub 11 Add to collection

正文以来源语言显示。

参考来源

← 返回每日简报