Reward-balancing methods normalize RL reward functions to enable greedy optimal policies, reformulated as optimal control with stochastic sampling for uncertainty and shown to improve performance in MPC simulations.
Title resolution pending
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
-
On Reward-Balancing Methods for Reinforcement Learning
Reward-balancing methods normalize RL reward functions to enable greedy optimal policies, reformulated as optimal control with stochastic sampling for uncertainty and shown to improve performance in MPC simulations.