Two-sided assembly lines are typically found in the production of large-sized products, such as automobiles and trucks. Two-sided assembly lines use both (left and right) sides of the line in parallel. Two-sided Assembly Line Balancing Problems (TALBP) is classified in an NP-Hard class; therefore it is difficult to obtain an optimal solution when the problem size increases. The objective of this research was to evaluate the performance of a new evolutionary method called Particle Swarm Optimization with Negative Knowledge (PSONK) and Local Search. This approach was applied to multi-objectives on mixed-model TALBP. Four objectives were considered including minimum number mated-stations, minimum number workstations, minimum work relatedness and minimum workload balance between workstations. The performance of PSONK was compared with COMSOAL, Non-dominated Sorting Genetic Algorithm II (NSGA-II), Memetic Algorithm (M-NSGA-II), COIN, and Discrete Particle Swarm Optimization (DPSO). The experimental results show that PSONK gives the best solutions for large-sized problems.