| Summary: | A real world environment is often partially observable by the agents either because of noisy sensors or incomplete perception. Moreover, it has continuous state space in nature, and agents must decide on an action for each point in internal continuous belief space. Consequently, it is convenient to model this type of decision-making problems as Partially Observable Markov Decision Processes (POMDPs) with continuous observation and state space. Most of the POMDP methods whether approximate or exact assume that the underlying world dynamics or POMDP parameters such as transition and observation probabilities are known. However, for many real world environments it is very difficult if not impossible to obtain such information. We assume that only the internal dynamics of the agent, such as the actuator noise, interpretation of the sensor suite, are known. Using these internal dynamics, our algorithm, namely Kalman Based Finite State Controller (KBFSC), constructs an internal world model over the continuous belief space, represented by a finite state automaton. Constructed automaton nodes are points of the continuous belief space sharing a common best action and a common uncertainty level. KBFSC deals with continuous Gaussian-based POMDPs. It makes use of Kalman Filter for belief state estimation, which also is an efficient method to prune unvisited segments of the belief space and can foresee the reachable belief points approximately calculating the horizon N policy. KBFSC does not use an “explore and update” approach in the value calculation as TD-learning. Therefore KBFSC does not have an extensive exploration-exploitation phase. Using the MDP case reward and the internal dynamics of the agent, KBFSC can automatically construct the finite state automaton (FSA) representing the approximate optimal policy without the need for discretization of the state and observation space. Moreover, the policy always converges for POMDP problems.
|
|---|
