Many variants of particle swarm optimization (PSO) have been proposed to improve convergence accuracy in applications to complex multimodal or real-world optimization problems, but this is at the price of an increase in the number of function evaluations. To deal with this problem, this paper proposes a single-vector PSO (SVPSO) based on a competition mechanism and an adaptive random adjustment strategy. First, to reduce the probability of particles falling into local optima, a collision random adjustment mechanism is employed to maintain the density of the population. Second, a leadership competition mechanism is used to balance exploitation and exploration in the search process by enlarging the search area dynamically. Third, a population-adaptive migration strategy is used to dispatch some particles to a new area when the population as a whole cannot achieve better fitness, which provides a powerful way to avoid premature convergence. Together with these methods, a single-vector structure for particles is adopted. The proposed SVPSO is evaluated on 16 benchmark functions and 12 real-world engineering problems in comparison with five state-of-the-art PSO variants. Experimental results and statistical analysis show that the proposed SVPSO performs better than the other algorithms in the majority of cases, especially with regard to accuracy and efficiency when applied to complex multimodal functions and real-world constrained optimization problems.