This paper considers the locally and globally optimal solutions of the global optimisation problem whose objective function is a max-plus vector-valued function and constraint function is a real affine function. The formulas about global optimisation in the existing work are further explained and simplified. The necessary and sufficient conditions for the existence and uniqueness of locally optimal solutions of a single objective are established, which are then used to deduce these of globally optimal solutions. Furthermore, the general formulas of locally and globally optimal solutions are presented, respectively. The local optimisation is then applied to solve the load distribution problem of distributed systems with no globally optimal solution, and the optimal allocation scheme is proposed to complete the overall task at the earliest time. The proposed method is constructive, and the obtained results are illustrated by the numerical examples.
INTRODUCTIONMax-plus linear systems can describe some nonlinear timeevolution systems with synchronisation but no concurrency, such as manufacturing systems, flow shop scheduling, traffic managements, communication networks (see, e.g. refs.[1-4]).Many researches have been made on control and optimisation of max-plus linear systems (see, e.g. refs. [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22]).Global optimisation is to find the global minimiser of a function or a set of functions over a given set, which has been a basic tool in all areas of engineering, medicine, economics, and so on (see, e.g. refs. [23, 24]). The global optimisation of maxplus linear systems has been considered in refs. [25] and[26], whose objective function is a max-plus vector-valued function and constraint function is a real affine function. This paper will further address the global optimisation problem considered in ref. [26]. The global optimisation problem involves some operations between infinity elements and real numbers, which are not clearly defined in ref. [26]. This paper will explain these operations definitely, and simplify the formula of the greatest lower bound and the discriminant of the existence of globally optimal solutions given in ref. [26]. In addition, the zero coefficient in the constraint function is neglected when it discusses the uniqueness of globally optimal solutions (see refs. [26] and [27]), and a sufficient condition for the uniqueness is presented.