Robot Control Using Stored, Generated, and Learned Action Plans
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robots are unable to adaptively perform operations based on their surroundings, leading to potential inefficiencies in task execution.
Innovation Solution
A robot equipped with a processor that acquires target and context information to select from pre-stored, context-generated, or learned action plans, and uses this information to control its operations, including utilizing a look-up table and AI models to generate or update action information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If robots perform operations according to predetermined orders, then the control logic is simple and easy to implement, but the robots are not able to adaptively perform operations according to the situations of the surroundings, leading to deterioration of job efficiency
Solution Approach 1:
The patent segments the control system into three distinct modules: a predetermined order execution module for basic tasks, a context information acquisition module for sensing environmental conditions, and a behavior selection module for adaptive decision-making. This segmentation allows the system to maintain simple predetermined operations while adding adaptive capabilities through modular components, resolving the contradiction between adaptability and control complexity.
Solution Approach 2:
The patent implements preliminary action by pre-storing multiple behavior sequences (patterns) in the memory unit, each corresponding to different situational contexts. The processor acquires context information in advance and selects the appropriate pre-stored behavior sequence based on the current situation, allowing the robot to adapt to environmental changes without requiring complex real-time decision algorithms, thus balancing adaptability with manageable system complexity.
2Productivity
If robots use predetermined action sequences, then the control implementation is straightforward, but the efficiency of job execution deteriorates when facing varying situational contexts
Solution Approach 1:
The patent applies dynamics by making the behavior selection adaptive rather than static. The processor dynamically selects from multiple pre-stored behavior sequences based on real-time context information acquisition. This allows the robot to maintain efficient predetermined execution for each specific situation while adapting the overall behavior to match environmental conditions, thereby improving both productivity and situational responsiveness.
Solution Approach 2:
The patent implements feedback through the context information acquisition unit that continuously monitors environmental conditions and feeds this information back to the processor. The processor uses this feedback to select the most appropriate pre-stored behavior sequence, creating a closed-loop control system that improves job execution efficiency by adapting to actual situational contexts rather than rigidly following predetermined sequences regardless of conditions.
Data Source
Figure 1~2
Figure 3~4a
Figure 4b
AI summary
A robot includes a driver, and a processor configured to: based on an occurrence of an event for performing an operation, acquire target information and context information related to the robot, based on at least one of the target information or the context information, select an action plan, acquire action information based on the action plan, and control the driver such that an operation corresponding to the target information is performed based on the action information. The action plan is selected as a first action plan of performing an operation according to first action information stored in advance in the memory, a second action plan of performing an operation according to second action information generated based on the acquired target information and the acquired context information, and a third action plan of performing an operation according to third action information learned based on an operation pattern of the robot.