Artificial Memory System for Predicting Human Operator Actions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for assisting human operators in dynamic environments, such as vehicles and aircraft, face challenges in handling and analyzing the increasing amount of data from sensors and actuators, particularly in predicting and anticipating human actions in real-time, especially in unknown or untested situations.
Innovation Solution
A Perceptual/Cognitive architecture based on a Memory Prediction Framework with an enhanced artificial hierarchical memory system, utilizing neural networks like Hierarchical Temporal Memory (HTM) for abstraction, generalization, and learning, which processes sensor data to understand and predict human intentions and vehicle behaviors, incorporating preprocessing stages for feature extraction and context awareness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of sensors and actuators is increased to monitor and control more components, then the monitoring coverage and control precision are improved, but the amount of data to be handled and analyzed increases significantly
Solution Approach 1:
The patent extracts only the essential and relevant features from the large volume of sensor data using feature extraction techniques. The system identifies and extracts key characteristics that are most relevant for predicting human actions, discarding redundant information and reducing the data volume that needs to be processed further.
Solution Approach 2:
The patent transforms the high-dimensional sensor data into a lower-dimensional representation by mapping sensor readings onto manifolds and extracting intrinsic features. This dimensional reduction maintains the essential information needed for prediction while significantly reducing the data volume for handling and analysis.
2Productivity
If traditional data processing methods are used to analyze sensor data, then the system structure is simple, but the system cannot effectively predict human actions in real-time, especially in unknown situations
Solution Approach 1:
The patent implements dynamic adaptation mechanisms where the system continuously learns from new data and adjusts its prediction models in real-time. The neural networks are trained online to adapt to unknown situations and changing patterns of human behavior, enabling the system to improve its prediction capability dynamically rather than relying on static pre-programmed rules.
Solution Approach 2:
The patent introduces intermediate processing layers including manifold learning, feature extraction modules, and hierarchical temporal memory structures that mediate between raw sensor data and final predictions. These intermediary components transform complex sensor inputs into meaningful representations that facilitate real-time prediction while managing system complexity through modular architecture.
3Loss of information
If all sensor data is processed and stored in detail, then the information completeness is improved, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary feature extraction and data transformation before the main prediction process. By pre-processing sensor data to extract essential features and organize information in meaningful structures, the system reduces the computational burden during real-time prediction while maintaining information completeness.
Solution Approach 2:
The patent divides the data processing task into multiple segments: initial feature extraction, manifold learning, hierarchical temporal processing, and final prediction. This segmentation allows the system to process different aspects of data at appropriate levels of detail, maintaining information completeness where needed while reducing processing time for routine operations.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This invention relates to an artificial memory system and a method of continuous learning for predicting and anticipating human operator´s action as response to ego-intention as well as environmental influences during machine operation. More specifically the invention relates to an architecture with artificial memory for interacting with dynamic behaviors of a tool and an operator, wherein the architecture is a first neural network having structures and mechanisms for abstraction, generalization and learning, the network implementation comprising an artificial hierarchical memory system.