Genetic Algorithm And Its Inverted Pendulum

Genetic algorithm is a global optimization algorithm which is based on natural selection and genetic mechanism.As genetic algorithm has the advantages in self-adaptive, high degree of parallelism and robustness,it has been applied in complex engineering domains widely.Though this algorithm achieves great effects in lots of problems, many deficiencies are exposed along with its superiority, such as premature convergence, bad in local searching etc.In this thesis, firstly, to aim at the blind search for cross-individual in the solution space, an improved instructed-crossover genetic algorithm based on gradient information is proposed.It executed crossover operation via choosing special individuals from the set range of the negative gradi-ent of the objective individuals which are got from the current popula-tion.The simulation on typical test functions indicates that the proposed algorithm can improve greatly the efficiency and the precision in search-ing the optimum value.Secondly, according to the limitation of the present evolutionary al-gorithm in solving constrained optimization, a new genetic algorithm for solving constrained optimization is proposed.It redesigned the fitness function for feasible and infeasible solution separately.Through the new function,we can manipulate constraint conditions easily.The simulation results of 13 test functions verified the effectiveness of the proposed al-gorithm.Finally, using the improved genetic algorithm, made an off-line op-timization for the controller parameters of double inverted pendulum.The results show that the system can get better performance on stabil-ity,overshoot and response speed for system response.