Adaptive pursuit-evasion under adversarial confrontations-part II: Risk-averse and efficient pareto strategies in linear-quadratic differential games

摘要

Adaptive decisions and risk-bearing actions for a linear-quadratic class of stochastic multi-player pursuit-evasion games with performance robustness against uncertain environments and adversarial confrontations are proposed with major theoretical advances and contributions as follows: i) an effective utilization of performance-measure statistics to directly address other higher-order uncertainties such as variance, skewness, flatness, etc. beyond the tradition of statistical averaging of probabilistic performance distributions; ii) a risk-value model for tradeoffs between performance values and risks to anticipatively ensure closed-loop performance robustness without relying on the intensively computational Monte Carlo performance dispersion analysis; and iii) risk-averse and efficient Pareto decision strategies to effectively handle adversarial jamming and confrontations from competing teams.