Soft-Computing Supervision for Dynamical Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing control design methods for dynamical systems, such as UAVs, face issues with hard discrete control law switching leading to instability and frequent mode changes due to conflicting control objectives, and fail to learn and adapt over time.
Innovation Solution
A method using distributed soft computing to receive state inputs from multiple dynamical systems, generate weights for preset control objectives, and transmit command signals to reduce mode switching and stabilize control, employing soft computing methods like fuzzy logic and neural networks to dynamically adjust control priorities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If hard discrete control law switching is used to manage conflicting control objectives, then the system can respond to changing conditions, but the system experiences frequent mode switching and instability
Solution Approach 1:
The patent applies dynamics by transitioning from fixed, discrete control laws to adaptive control laws that dynamically adjust based on system state. The neural network learns optimal control strategies online, allowing the system to adapt to changing conditions while maintaining stability through continuous, smooth adjustments rather than abrupt discrete switches.
Solution Approach 2:
The patent changes the parameter of control law selection from discrete fixed modes to continuous adaptive parameters learned by neural networks. The control strategy evolves by modifying network weights and biases based on system performance, enabling smooth transitions between control modes and eliminating the instability caused by hard switching.
2Ease of manufacture
If fixed algorithms are used for control law switching, then the switching logic is simple to implement, but the system cannot learn from experience and produces repeated instability
Solution Approach 1:
The patent implements self-service by enabling the control system to automatically learn and improve from its own operational experience. The neural network performs online learning, adjusting its parameters based on system performance feedback, allowing the system to self-optimize control strategies without external intervention while maintaining implementation feasibility through modular architecture.
3Ease of manufacture
If weighted combinations of control objectives are used with fixed weight prioritization, then the control design is straightforward, but the system cannot adequately capture prevailing system conditions
Solution Approach 1:
The patent makes the weight prioritization dynamic by using neural networks to adaptively determine weights based on current system conditions. The network learns optimal weight assignments online, allowing the control system to accurately capture prevailing conditions while maintaining design simplicity through the modular neural network architecture that integrates with existing control frameworks.
Data Source
AI summary
A method to supervise a local dynamical system having multiple preset control objectives and operating in conjunction with other dynamical systems. The method includes receiving state input from dynamical systems in an environment at a distributed soft computing level, generating weights and applying the weights to the preset control objectives using soft computing methods to form weighted control objectives. The weights are computed based on the received state input. The method also includes generating a command signal for the local dynamical system based on the weighted control objectives and transmitting the command signal to a controller in the local dynamical system.