Learning of Human Control Strategy: Continuous and Discontinuous
This chapter explores the machine learning techniques and statistical analysis towards abstracting models of human control strategy. It proposes that an efficient continuous learning framework for modeling human control strategy that combines cascade neural networks and node-decoupled extended Kalman filtering. The chapter discusses the cascade learning towards abstracting human control strategy models from experimental control strategy data. It aims to develop a stochastic, discontinuous modeling framework for modeling discontinuous human control strategies. This approach models different control actions as individual statistical models, which, together with the prior probabilities of each control action, combine to generate a posterior probability for each action, given the model inputs. The proposed algorithm models possible control actions as individual statistical models. During run-time execution of the algorithm, a control action is then selected stochastically, as a function of both prior probabilities and posterior evaluation probabilities.