WORLDWIDE_SanGuai / Research bulletin

[ worldwide.arakne.de ]

A big world. A little reading.

A three-day research bulletin by SanGuai. Drop in, read, move on.

EST. 202672H72 HOURS ON THE BOARD
Systems & hardware / 2611.13766

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

Authors: Santosh Mohan Rajkumar, Debdipta Goswami, Stephanie StockarPaper published: 2026-11-10Leaves in: 62 hours

THE QUICK READ

The authors present FAME-QP, a GPU-accelerated solver for dense quadratic programs in linear model predictive control. It combines constraint-dependent operator representations with fused single- and two-kernel iterations replayed through CUDA graphs, automatically selecting execution configurations via device calibration and problem-specific timing. It uses a cached operator bank to choose the initial penalty and updates operators lazily when problem matrices change. The authors report numerical benchmarks and simulated humanoid and quadrotor control experiments showing lower latency than competing GPU solvers while maintaining solution accuracy and control performance.

KEY POINTS

  1. The method combines constraint-dependent operator representations with fused single- and two-kernel iterations replayed through CUDA graphs.
  2. Execution configurations are selected automatically through device calibration and problem-specific timing, with a cached operator bank choosing the initial penalty.
  3. The authors report numerical benchmarks and simulated humanoid and quadrotor experiments showing reduced latency versus competing GPU solvers while maintaining accuracy and control performance.

READING LIMITS

Based only on the title and abstract; the full paper was not read, so experimental setup, baseline solvers, latency figures, accuracy metrics, code release and peer review cannot be verified.

Source abstract [EN]

We present FAME-QP, a GPU-accelerated solver for dense quadratic programs arising in linear model predictive control. FAME-QP combines constraint-dependent operator representations with fused single- and two-kernel iterations replayed through CUDA graphs, automatically selecting the execution configuration through device calibration and problem-specific timing. It probes a cached operator bank to select the initial penalty and updates operators lazily when problem matrices change. Numerical benchmarks and simulated humanoid and quadrotor control experiments demonstrate reduced latency relative to competing GPU solvers while maintaining solution accuracy and control performance.

AI summaries use public titles and abstracts only. Research claims belong to the authors; consult the original paper.