A self-organizing fuzzy PID controller (SFPID), which is composed of a PID type fuzzy controller (FPID) and a self-organizing adjustment mechanism, is proposed to provide the balance of the control precision and dynamic performance. A normalized acceleration introduced into the self-organizing mechanism, as an indicator of system response speed, is used to identify the system error and various stages in whole control process. The integral and derivative coefficients of FPID are adjusted in an online manner with the aid of the normalized acceleration. For avoiding complex adjustment of parameters in the design of controllers and attaining optimal control properties, a chaos particle swarm optimization algorithm (CPSO) is applied to optimizing the SFPID parameters, in which chaotic searching is employed to overcome the particle's ‘inertia’, and improve the global search performance and local convergence. The results of the application to typical controlled objects and LED source-meter measuring system show that the SFPID exhibits good dynamic performance and robustness to the uncertainty of system parameters, which can effectively eliminate steady-state deviations with minor overshooting.