Abstract. The paper considers the optimization problem of packing different solid spheres into containers of types: a cuboid, a sphere, a right circular cylinder, an annular cylinder, and a spherical layer. The radii of spheres are assumed to be variables. This allows us to propose a new technique to derive initial points belonging to the feasible region of the problem, as well as to carry out a non-exhaustive search of local extrema, using a modification of the jump algorithm (JA), which implements a continuous transition from one local minimum to another with a better value of the objective. A reduction of the solution space dimension of the problem and rearrangements of sphere pairs allow improving the objective function value. The results obtained are compared with benchmark ones.

Keywords: packing, sphere packing, non-convex optimization problem, jump algorithm.