Multi-Solver Simulation Controller for Dynamic Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex dynamic systems require accurate and detailed modeling, which becomes increasingly time-consuming and effort-intensive as systems become more complex, especially when dealing with hybrid dynamics spanning multiple domains and requiring communication between solvers to maintain coherent simulations.
Innovation Solution
Implementing a simulation controller to schedule and control the execution of multiple solvers, allowing them to exchange information during minor steps, rather than only at major steps, to improve approximation convergence and synchronization, even when solvers have different step sizes or execute on different cores/processors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple solvers are used to simulate complex dynamic systems, then simulation accuracy and convergence rates improve, but computational complexity and coordination overhead increase
Solution Approach 1:
The patent divides the complex dynamic system into multiple subsystems, each simulated by a dedicated solver. This segmentation allows each solver to handle specific portions of the system independently, improving simulation accuracy through specialized processing while managing computational complexity through modular organization.
Solution Approach 2:
The patent introduces a simulation controller as an intermediary that coordinates communication between multiple solvers. The controller manages data exchange, synchronizes simulation steps, and handles coordination overhead, allowing multiple solvers to work together efficiently without direct complex interactions between them.
2Productivity
If solvers exchange information only at major steps, then computational overhead is reduced, but approximation accuracy and convergence rates deteriorate
Solution Approach 1:
The patent implements periodic information exchange between solvers at both major steps and minor steps. This periodic communication pattern allows solvers to share intermediate results more frequently, improving approximation accuracy and convergence rates while maintaining computational efficiency through structured, periodic interactions rather than continuous communication.
Solution Approach 2:
The patent enables solvers to exchange information at minor steps before reaching major steps. This preliminary action allows solvers to share intermediate calculations and predictions early in the simulation process, improving overall approximation accuracy by correcting trajectories before final major step computations.
3Power
If solvers execute on different cores/processors with different step sizes, then parallel processing efficiency improves, but synchronization difficulty and computational errors increase
Solution Approach 1:
The patent implements dynamic step size management where each solver can operate with different step sizes appropriate to its subsystem, enabling efficient parallel processing. The simulation controller dynamically coordinates these different step sizes, adjusting synchronization points and interpolation strategies to maintain reliability despite varying computational rates across processors.
Solution Approach 2:
The patent allows solvers to use different numerical parameters including step sizes, tolerance levels, and integration methods optimized for their specific subsystems. The simulation controller manages these parameter variations, ensuring that differences in solver parameters do not compromise overall simulation reliability through appropriate coordination and validation.
Data Source
AI summary
Two or more solvers used in a simulation of a dynamic system may exchange information during minor steps of at least one of the solvers. The information may be exchanged one way—that is, a first solver sending the information to a second solver after the second solver's minor step, or it may be exchanged two ways (or multiple ways in a case with more than two solvers). If the solvers are instances of the same type of solver, they may exchange information at each minor step.