Digital Twin Learning Device for Mobile Cyber-Physical Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous systems, such as autonomous vehicles, struggle to adapt their machine learning algorithms to the specific physical features and changes over time, leading to potential trajectory errors in operational conditions, as these algorithms are typically trained in controlled test environments and not updated to account for real-world variations.
Innovation Solution
A learning cyber-physical system that combines conventional offline learning with simulated learning using data from the system's environment and internal sensors, allowing for regular updates and adaptation through a digital twin simulation, enabling the machine learning algorithm to adjust its behavior based on the system's evolving features and operational conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning algorithms are trained in controlled test environments, then learning safety and control are improved, but adaptability to real-world variations and physical feature evolution deteriorates
Solution Approach 1:
The patent creates a digital twin copy of the physical system that replicates its physical features and behavior. This digital twin is used to simulate real-world conditions and train machine learning algorithms, allowing safe reproduction of physical system characteristics without actual physical risk. The digital twin captures evolution of physical features over time, enabling the learned models to adapt to real-world variations while maintaining training safety.
2Adaptability or versatility
If learning algorithms are updated to account for physical feature evolution, then adaptability to operational conditions is improved, but system complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-training machine learning algorithms on simulated data from digital twins that represent various operational conditions and physical feature evolutions. This preliminary training equips the algorithms with knowledge of potential variations before they occur in reality, enabling continuous adaptation without requiring complex real-time learning systems. The digital twin simulates future states and edge cases in advance, reducing the need for complex adaptive mechanisms during actual operation.
Data Source
AI summary
A learning device intended to be included in a mobile cyber-physical system provided with actuators, the device comprising at least one perception sensor for perceiving the external environment of the system, at least one internal sensor able to provide information concerning the state of the system, a first learning unit configured to render a perception of the environment from the data acquired by the at least one perception sensor, a second learning unit configured to control the actuators, a generator for generating simulation scenarios of the system in its environment controlled by the first learning unit and the second learning unit, a scenario simulator and a virtualization platform for simulating the behavior of a digital twin of the system in the scenarios simulated by the generator and for adapting the parameters of the second learning unit in order to control the system so that it adapts to its environment.