MSc Student of CINVESTAV Guadalajara
Karim Luna
I work on control, optimization, and approximation of dynamic systems, applying hardware-aware algorithms to world problems, for example in the study of how resource-constrained units can cooperate toward goals none of them could reach alone. My current frontiers are approximation of dynamical systems, world models, and multi-agent control.
SELECTED WORK
Double Inverted Pendulum: SAC Swing-Up + Policy Compression
SAC swing-up controller distilled to an INT8 policy small enough to run on an STM32H7.
PyTorch · MuJoCo · STM32H7 · C
Ground Robot MRTA: Auction Baselines vs. MARL
Multi-robot task allocation in MuJoCo + PettingZoo, comparing auction-based allocation against learned coordination.
MuJoCo · PettingZoo · Python