Multitask LQG control via history-dependent lifting to LQR yields generalization bounds tied to bisimulation heterogeneity and reduces policy gradient variance proportionally to the number of training tasks.
On the lack of gradient domination for linear quadratic Gaussian problems with incomplete state information,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Multitask LQG Control: Performance and Generalization Bounds
Multitask LQG control via history-dependent lifting to LQR yields generalization bounds tied to bisimulation heterogeneity and reduces policy gradient variance proportionally to the number of training tasks.