Robot Recognition Model Updating for Continuous Operation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robotic systems installed in environments that require continuous operation, such as plants, face challenges in performing tasks efficiently due to the need for trial-and-error learning to achieve a high success rate, which is time-consuming and difficult in real-world settings.
Innovation Solution
An information processing apparatus that acquires a learning model from an external server and uses it for incremental learning in the robotic system, allowing the model to be updated without requiring the robot to perform tasks in a trial-and-error manner, by collecting learning data and replacing the existing model when the new model meets predetermined performance levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a robot performs trial-and-error learning in a real-world environment to achieve a high success rate, then the robot can learn a model effectively, but it takes much time and is difficult to implement in continuous operation environments
Solution Approach 1:
The patent applies preliminary action by performing trial-and-error learning in a simulation environment before deploying the robot in the real world. The learning process is completed in advance in the virtual space, creating a pre-trained model that can be directly applied to real-world operations without requiring extensive trial-and-error time in the actual environment.
Solution Approach 2:
The patent uses copying by creating a virtual copy of the real-world environment through simulation. The robot learns in this copied virtual environment, and the learned model is then transferred to the real robot. This allows learning to occur in a risk-free simulated space that mirrors real-world conditions without the time constraints and risks of actual physical trial-and-error.
2Adaptability or versatility
If a robot performs incremental learning in a real-world environment to improve performance, then the model can be continuously improved, but it requires trial-and-error operations that are difficult in continuous operation settings
Solution Approach 1:
The patent introduces an intermediary simulation environment that mediates between the real-world robot operations and the learning process. Instead of performing trial-and-error learning directly in the real world, the simulation acts as an intermediary where learning can occur freely, and the results are transferred back to improve the real robot's performance without disrupting continuous operations.
3Reliability
If a robotic system performs trial-and-error learning to achieve high task success rate, then learning effectiveness is improved, but productivity decreases due to time consumption
Solution Approach 1:
The patent resolves this contradiction by performing the time-consuming learning action in advance in a simulation environment. The trial-and-error learning is completed before the robot needs to operate productively in the real world, so the full learning process does not interfere with ongoing productive operations.
Solution Approach 2:
The patent implements periodic action by alternating between simulation-based learning phases and real-world operational phases. During periodic learning intervals in simulation, the model is improved without affecting real-world productivity. During operational phases, the robot uses the pre-trained model to maintain high productivity while achieving high task success rates.
Data Source
AI summary
The invention provides an information processing apparatus comprising a first acquiring unit which acquires, from an external apparatus which can communicate with the information processing apparatus via a network, a first learning model which outputs recognition information in response to input of image information; a learning unit which causes the first model to learn a result of control by the information processing apparatus using recognition information that is output from a second learning model in response to input of the image information in an execution environment; and an output unit which causes the first learning model learned by the learning unit to output recognition information by inputting image information in the execution environment to the first learning model.