Autonomous Control Learning Through Self-Generated Sensor Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing artificial intelligence methods require pre-defined training data, which is error-prone and complex, and cannot correct errors or adapt to unknown objects, leading to inefficient and unreliable autonomous control of devices.
Innovation Solution
A method for autonomously controlling a device through randomized actuation of effectors, using internal and external sensors to log interrelationships between effector actions and environmental properties, converting output triplets into target triplets, and iteratively learning and storing instructions for device control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pre-defined training data is used for AI learning, then the learning process can be structured and guided, but the system becomes error-prone and cannot adapt to unknown objects or correct errors
Solution Approach 1:
The system performs self-exploration and self-learning by autonomously generating actions and observing outcomes without requiring pre-defined training data. The device learns through its own experiences in the real environment, making the system self-sufficient and adaptable to unknown situations while maintaining reliability through continuous self-correction
Solution Approach 2:
The learning system transitions from static pre-defined training data to dynamic real-time learning. The system continuously updates its knowledge base through ongoing exploration and observation, allowing it to adapt to changing environments and unknown objects while maintaining structured learning through its exploration framework
2Adaptability or versatility
If comprehensive training data is provided to cover all possible situations, then the system can handle diverse scenarios, but the complexity and effort of data preparation increases significantly
Solution Approach 1:
Instead of requiring external preparation of comprehensive training data, the system generates its own training data through autonomous exploration. The device independently discovers scenarios and learns from real-world interactions, eliminating the complex data preparation process while maintaining comprehensive scenario coverage through continuous exploration
Solution Approach 2:
The system performs preliminary exploration actions to discover the environment and generate training data before formal learning begins. By proactively exploring the real world and logging outcomes, the system prepares its own training data dynamically, reducing external data preparation complexity while ensuring comprehensive scenario coverage
3Productivity
If the system continuously learns and updates its knowledge base, then the device improves its performance over time, but the computational resources and processing time increase
Solution Approach 1:
The system performs continuous learning through ongoing exploration and observation without interruption. By continuously generating actions, observing outcomes, and updating the knowledge base in real-time, the system improves productivity over time while minimizing idle processing time through uninterrupted learning cycles
Solution Approach 2:
The system alternates between exploration phases and utilization phases, periodically updating the knowledge base. During exploration, new data is collected and learned; during utilization, the learned knowledge is applied. This periodic cycle balances continuous improvement with efficient processing, reducing overall processing time while maintaining productivity gains
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a method for the autonomous control of a device, wherein the device generally moves in a physical real world and no longer needs to be restricted with regard to its physical design. The method offers the advantage that the device performs autonomous learning and continuously improves the learned knowledge or behavior. In general, it overcomes the disadvantage of the prior art, which requires the creation of training data through selection and interpretation, as is the case with conventional artificial intelligence methods. The method is generally universally applicable, and the device learns autonomously, constantly correcting its own knowledge base. Furthermore, a device configured to execute the method is proposed, as well as a system arrangement comprising several of the proposed devices.Furthermore, a computer program product and a computer-readable storage medium are proposed, which execute the process steps or cause a computer to execute the process.