Convergence Of Finite Memory Q-learning For Pomdps And Near Optimality Of Learned Policies Under Filter Stability

Abstract

In this paper, for POMDPs, we provide the convergence of a Q learning algorithm for control policies using a finite history of past observations and control actions, and, consequentially, we establish near optimality of such limit Q functions under explicit filter stability conditions. We present explicit error bounds relating the approximation error to the length of the finite history window. We establish the convergence of such Q-learning iterations under mild ergodicity assumptions on the state process during the exploration phase. We further show that the limit fixed point equation gives an optimal solution for an approximate belief-MDP. We then provide bounds on the performance of the policy obtained using the limit Q values compared to the performance of the optimal policy for the POMDP, where we also present explicit conditions using recent results on filter stability in controlled POMDPs. While there exist many experimental results, (i) the rigorous asymptotic convergence (to a

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keykara2021convergence

Related papers

• Finite-time Analysis Of Natural Actor-critic For Pomdps (2022)0.00
• Model-based Learning Of Near-optimal Finite-window Policies In Pomdps (2026)0.00
• Q-learning For Mdps With General Spaces: Convergence And Near Optimality Via Quantization Under Weak Continuity (2021)0.00
• Finite-state Controllers For (hidden-model) Pomdps Using Deep Reinforcement Learning (2026)0.00
• Computationally Efficient PAC RL In Pomdps With Latent Determinism And Conditional Embeddings (2022)0.00
• A Policy Gradient Method For Confounded Pomdps (2023)0.00
• Hidden Markov Model Estimation-based Q-learning For Partially Observable Markov Decision Process (2018)5.84
• Globally Convergent Policy Search Over Dynamic Filters For Output Estimation (2022)0.00