There are many requirements, issues and complexities that engineering design problems must take into account, and these engineering items are often treated as constraints in optimization problems. The constraint optimization problems (COPs) are the most popular type of problems, therefore, optimization algorithms for the COPs have been attracted more and more attention. Particle swarm optimization is one of the most popular algorithms for solving the COPs due to the simplicity and efficient convergence performance, and various types of PSO are actively developed in recent years. However, despite the popularity, they still have limitations in terms of search efficiency. Notably, it takes a lot of calculation costs to obtain the optimal solution on the boundary of constraints or the complicated feasible area. To overcome the difficulties, this paper presents a novel particle swarm optimization algorithm named independent 2-group particle swarm optimization (I2GPSO). I2GPSO is based on the following ideas - a constraint handling method, a novel structure of particles and a novel local search operator. The constraint handling method uses the existing penalty function method. The structure of particles defines two particle groups that have original roles, efficiently enabling PSO to search globally and locally. The local search operator is introduced into one group and enables particles to search near the boundary of constraints efficiently or candidates of the optimal solution. These novel approaches effectively reinforce the optimization efficiency of the PSO algorithm. The optimization capability and character of I2GPSO is illustrated in 11 benchmark problems. The results are compared with other state-of-the-art PSOs, and it is shown that the proposed algorithm possesses competitive search efficiency.