Microfluidic Biochip Control Logic With DRL Pattern Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing control logic designs for continuous microfluidic biochips are inefficient due to the lack of optimization in the allocation order between control patterns and multi-channel combinations, leading to redundant resource usage and increased costs.
Innovation Solution
A Deep Reinforcement Learning (DRL) based control logic design method that calculates a multi-channel switching scheme and allocates control patterns to minimize the number of time slices and control valves used, employing integer linear programming and double deep Q-networks for optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional control logic design methods are used, then the design process is simple, but the number of time slices and control valves increases leading to redundant resource usage
Solution Approach 1:
The patent changes the optimization parameters from traditional approaches to specifically target the number of time slices and control valves. By formulating the problem with explicit parameters for time slice count and valve count, the DRL agent can learn optimal control logic that minimizes these specific resource quantities while maintaining functional requirements.
Solution Approach 2:
The patent replaces traditional mechanical/control-based optimization methods with a deep reinforcement learning system. The DRL agent learns optimal control logic through simulated training, substituting conventional iterative design and manual optimization with an intelligent learning system that automatically discovers resource-efficient control strategies.
2Quantity of substance
If the number of control valves is reduced through multiplexing, then resource usage decreases, but the allocation order between control patterns and multi-channel combinations becomes suboptimal
Solution Approach 1:
The patent introduces dynamic optimization through DRL, where the control logic is not statically determined but adaptively learned. The system dynamically adjusts control patterns and multi-channel combinations based on learned policies, allowing optimal allocation that simultaneously minimizes valve count while maintaining high execution efficiency for biochemical applications.
Solution Approach 2:
The patent performs preliminary training of the DRL agent in a simulated environment before deployment. Through extensive pre-training on various biochemical application scenarios, the agent learns optimal control strategies in advance, enabling it to efficiently allocate control valves and patterns when deployed in real applications without requiring real-time complex calculations.
3Device complexity
If existing control logic optimization methods are used, then some parameters are improved, but the allocation order optimization is insufficient leading to redundant resources
Solution Approach 1:
The patent implements feedback mechanisms through the DRL training process, where the agent receives reward signals based on the efficiency of its control logic design. The feedback loop continuously refines the agent's policy, allowing it to learn from past decisions and progressively improve the allocation of control resources, eliminating redundant elements through iterative optimization.
Solution Approach 2:
The patent enables the control logic design system to optimize itself automatically without external intervention. The DRL agent independently analyzes control requirements, evaluates different allocation strategies, and determines optimal control patterns and multi-channel combinations, making the system self-sufficient in generating resource-efficient control logic.
Data Source
AI summary
A DRL-based control logic design method for continuous microfluidic biochips is provided. Firstly, an integer linear programming model is for effectively solving multi-channel switching calculation to minimize the number of time slices required by the control logic. Secondly, a control logic synthesis method based on deep reinforcement learning, which uses a double deep Q network and two Boolean logic simplification techniques to find a more effective pattern allocation scheme for the control logic.