Camera Module PID Parameter Optimization via Genetic Algorithm
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optical image stabilization (OIS) camera modules face challenges in finding PID parameters that satisfy desired response characteristics, requiring long modeling times and using algorithms with low processing speed and suboptimal solutions.
Innovation Solution
An apparatus and method utilizing a genetic algorithm with weight-gain elitism to rapidly generate PID child parameter entities, optimizing the PID parameters for image stabilization by applying weight-gain elitism to the parameter entities based on step response characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional algorithms are used to find PID parameters, then the system can find a solution, but the processing speed is low and the solution is not optimal
Solution Approach 1:
The patent replaces conventional iterative algorithms with a genetic algorithm that uses biological evolution metaphors (selection, crossover, mutation) to optimize PID parameters. This substitution enables parallel processing of multiple parameter sets simultaneously, dramatically improving processing speed while maintaining solution quality through fitness-based selection mechanisms.
Solution Approach 2:
The patent implements preliminary action by generating an initial population of PID parameter sets before the optimization process begins. These pre-generated parameter sets serve as the starting point for genetic operations, allowing the system to explore multiple solutions in parallel from the outset rather than sequentially searching for one solution at a time.
2Loss of time
If conventional modeling methods are used to find PID parameters, then a solution can be obtained, but the modeling time is long
Solution Approach 1:
The patent implements self-service by using the system's own output responses to automatically evaluate and optimize its PID parameters through the genetic algorithm. The system generates parameter sets, tests them against the plant model, evaluates fitness based on performance criteria, and iteratively improves parameters without external intervention, dramatically reducing modeling time.
Solution Approach 2:
The patent employs feedback mechanisms where the fitness evaluation of each PID parameter set is fed back into the genetic algorithm to guide the selection, crossover, and mutation operations. This closed-loop feedback enables rapid convergence to optimal parameters by continuously learning from previous iterations, significantly reducing the time required to achieve desired performance.
3Productivity
If the genetic algorithm processes all parameter entities equally, then the algorithm is simple, but the convergence speed is slow and error reduction is inefficient
Solution Approach 1:
The patent applies local quality by implementing differential weighting where parameter entities with better fitness values receive higher weights and thus greater influence on the next generation. Instead of treating all parameter sets equally, the system selectively amplifies the contribution of high-performing parameters through weight-gain elitism, accelerating convergence while maintaining algorithmic feasibility.
Solution Approach 2:
The patent utilizes parameter changes by dynamically adjusting the weights assigned to different parameter entities based on their fitness values. This parameter transformation converts the static genetic algorithm into an adaptive system where the selection pressure and evolutionary direction are continuously modified based on real-time performance feedback, improving convergence speed without excessive complexity.
Data Source
AI summary
An embodiment discloses an apparatus for controlling a camera module, including a parameter generation unit configured to generate a parameter entity group, calculate a fitness value of the parameter entity group based on a step response characteristic of an output signal, and generate an offspring entity by applying weight-gain elitism to a parameter entity having the smallest fitness value among parameter entities included in the parameter entity group, and a controller configured to generate a control signal by multiplying an error signal by a parameter.