Affective Interaction Based Decision-Making Approaches With Applications In Process Control

It is acknowledged that numerous decision-making tasks related with industrial process control engineering such as control performance assessment and multi-variable coordinating control are performed through human-computer interactions. In response to the problems that traditional interactive evolutionary computing approaches suffer limited searching ability and human’s strong subjectivity as well as the difficulties in comprehensive evaluation of objectives or attributes, a novel affective learning and evolution scheme consistent with human-computer interactive mechanism is explicitly proposed. Therein, a kind of stimulating response based affective computing models is constructed along with quantitative relations between affective space and human’s decision-making preferential membership functions, followed by providing affective learning strategies in terms of decision-making preferences with giving analysis of the algorithm’s complexity and proving the algorithm’s convergence; subsequently, the affective interactive decision making is described as a multi-objective/multi-attribute fuzzy mathematical programming problems based on evolutionary algorithms, human’s decision-making preferences are evolving during the human-computer interaction. Additionally, colored Petri nets’ realization techniques of agents are discussed in details, with the aid of the Petri nets’analysis methods, the affective interactive evolutionary decision-making technology theories are established. The main contribution could be addressed as follows.1. Taking advantage of affective space representations of hidden Markov chain models (HMM), frameworks of personality OCC (ORTONY, CLORE, COLLINS) models, as well as implications of a kind of human psychology stimulated model, a stimulated and transferring affective computing model (STAM) is proposed, which provides with a method to quantitatively computing human’s affective recognitions against continuous external persistent stimuli, establishing theoretical basis for the research work.2. Mathematical descriptions of affective decision-making preferences are established, along with quantitative relations between affective space and decision-making preferential membership functions as well as genetic algorithms based affective interactive decision-making preferences learning strategies. The algorithm’s time complexity and space complexity analysis are provide, along with proving the algorithm’s convergence. Experimental results show that the proposed learning approaches can help gradually grasp essentials in human’s affective preferences towards decision-making, reduce human’s subjective fatigue in the interactive process as well as make the decisions more scientific and objective.3. Consequently, decision-makers’knowledge and understanding deep about the problems can be responsible for qualities of final solutions in complex multi-objective/attribute interactive decision-making problems. Then, a multi-objective/attribute fuzzy mathematical programming solution is put forward, which can evolve the objective/attribute weight preference, with providing a new technical method for multi-objective/multi-attribute decision-making problems.4. An affective interaction based learning Agent (Agent-BDI)model is established, where colored Petri nets are employed to implement agents’ scheduling procedures. Based on colored Petri nets, overlay tree and CPN-Tools simulation analysis methods are employed to make analysis towards Agent with proving the proposed Agent model’s correctness, along with providing a software scheme for human affective interactive decision-making.5. The affective interactive decision-making methods are applied to control loop assessment and parameter tuning, as well as multi-variable coordinating control achievement, which are allegedly referred to as two typical interactive decision-making issues in process control field, consequently leading to satisfactory experimental results. In conclusion, the main innovative aspects of this thesis could be presented as follows. First, a kind of stimulus response mechanism affective computing models is proposed. Second, mapping relations between affective space and decision-making preferences are established, which results in GA based affective learning algorithms. Third, affective interaction based evolutionary decision-making algorithms are created under the interactive evolutionary computation framework. Affective interaction agents are realized by means of colored Petri nets with discussing the solution to process control engineering problems. The meaningful contribution of this thesis lies in the fact that the affective computing theories and technologies have been made inroads into solving human-computer interactive decision-making problems, which definitely illustrates the interdisciplinary characteristics.