Predictive Controller Solver Switching for Embedded MI-MPC
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Implementing mixed-integer model predictive control (MI-MPC) on embedded control systems with limited computational capabilities is challenging due to the complexity of solving mixed-integer optimization problems in real-time.
Innovation Solution
The approach involves using fast heuristic techniques to find feasible but potentially suboptimal solutions, and selectively employing strong and weak solvers based on the region of the state space, with disagreement regions identified using supervised learning algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If mixed-integer model predictive control (MI-MPC) is implemented on embedded control systems, then control accuracy and system optimization are improved, but computational complexity and resource requirements increase significantly
Solution Approach 1:
The patent segments the state space into multiple regions and assigns different solvers to different regions. This segmentation allows the system to use computationally intensive strong solvers only in critical regions while using lighter weak solvers in other regions, thereby maintaining control accuracy where needed while reducing overall computational complexity on embedded systems.
Solution Approach 2:
The patent applies local quality by using different solver strengths in different state space regions. Strong solvers with high computational power are deployed in regions requiring precise control, while weak solvers are used in regions where approximate solutions suffice. This localized approach optimizes the balance between control accuracy and computational resource usage.
2Measurement precision
If strong solvers are used to solve optimization problems, then solution accuracy is improved, but computational resource consumption and solving time increase
Solution Approach 1:
The patent implements dynamic solver selection based on the current state space region. The system dynamically switches between strong and weak solvers depending on the operational context, using strong solvers only when high accuracy is critical and weak solvers when computational resources are constrained. This dynamic adaptation resolves the contradiction between solution accuracy and computational resource consumption.
Solution Approach 2:
The patent changes the solver parameter (strength/computational power) based on the state space region. By adjusting the solver selection parameter dynamically according to the system state, the patent achieves high solution accuracy only when necessary, thereby reducing overall computational resource consumption while maintaining adequate performance across different operating conditions.
3Speed
If fast heuristic techniques are used, then computational speed is improved, but solution optimality deteriorates
Solution Approach 1:
The patent applies partial action by using fast heuristic techniques (weak solvers) in state space regions where approximate solutions are sufficient, while reserving exact optimization methods (strong solvers) for critical regions. This partial use of computationally intensive methods maintains solution optimality where needed while achieving faster computational speeds in non-critical regions, resolving the contradiction between speed and optimality.
Data Source
AI summary
A method for controlling a system by a controller comprises accepting a current state of the system and selecting, using a trained function of the current state, a solver from a set of solvers. The method further comprises solving an optimal control optimization problem using the selected solver to produce a current control input, such that for at least some different control steps, the predictive controller solves a formulation of the optimal control optimization problem with different solvers having different accuracies, requiring different computational resources, or both and submitting the current control input to the system thereby changing the current state of the system.