EVOLUTION OF DYNAMIC GAMES AND BACKWARD INDUCTION

As an efficient method of solving subgame-perfect Nash equilibrium,the backward induction is analyzed from an evolutionary point of view in this paper,replacing a player with a population and turning a game into a population game,which shows that equilibrium of a perfect information game is the unique evolutionarily stable outcome for dynamic models in the limit.

Powertrain Fuel Consumption Modeling and Benchmark Analysis of a Parallel P4 Hybrid Electric Vehicle Using Dynamic Programming

The goal of this work is to develop a hybrid electric vehicle model that is suitable for use in a dynamic programming algorithm that provides the benchmark for optimal control of the hybrid powertrain. The benchmark analysis employs dynamic programming by backward induction to determine the globally optimal solution by solving the energy management problem starting at the final timestep and proceeding backwards in time. This method requires the development of a backwards facing model that propagates the wheel speed of the vehicle for the given drive cycle through the driveline components to determine the operating points of the powertrain. Although dynamic programming only searches the solution space within the feasible regions of operation, the benchmarking model must be solved for every admissible state at every timestep leading to strict requirements for runtime and memory. The backward facing model employs the quasi-static assumption of powertrain operation to reduce the fidelity of the model to accommodate these requirements. Verification and validation testing of the dynamic programming algorithm is conducted to ensure successful operation of the algorithm and to assess the validity of the determined control policy against a high-fidelity forward-facing vehicle model with a percent difference of fuel consumption of 1.2%. The benchmark analysis is conducted over multiple drive cycles to determine the optimal control policy that provides a benchmark for real-time algorithm development and determines control trends that can be used to improve existing algorithms. The optimal combined charge sustaining fuel economy of the vehicle is determined by the dynamic programming algorithm to be 32.99 MPG, a 52.6% increase over the stock 3.6 L 2019 Chevrolet Blazer.

A Two-Level Hierarchical Markov Decision Model with Considering Interaction between Levels

Decision in reality often have the characteristic of hierarchy because of the hierarchy of an organization＇s structure. In this paper, we propose a two-level hierarchic Markov decision model that considers the interactions of agents in different levels and different time scales of levels. A backward induction algo- rithm is given for the model to solve the optimal policy of finite stage hierarchic decision problem. The proposed model and its algorithm are illustrated with an example about two-level hierar- chical decision problem of infrastructure maintenance. The opti- mal policy of the example is solved and the impacts of interactions between levels on decision making are analyzed.

A Syntactic Approach to Knowledge in Finite Games of Perfect Information

In this paper I present a syntactic approach to modeling the interactive knowledge of rationality in finite games of perfect information. This approach allows for a more transparent interpretation. In particular, we have the intuitive picture of viewing knowledge as the input and decisions as the output of a player＇s deliberation. This distinction is blurred in the semantic state-space approach.