Robot Path Generation Using Shape Estimation From Fewer Training Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating an action path for a robot require numerous training points to accurately follow a target object's shape, especially for curved surfaces, leading to a tradeoff between training accuracy and workload.
Innovation Solution
A path generation device that receives shape classifications of a target object, instructs the user on the number of training points needed for each classification, acquires positional information, estimates the target object's shape, and generates a path for the robot to follow, thereby reducing the number of training points required while increasing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If straight-line interpolation of positional information for training points is used to generate an action path, then the robot can follow a path, but many training points are needed to achieve good accuracy along the shape of the target object, especially for curved surfaces
Solution Approach 1:
The patent replaces the mechanical approach of using numerous discrete training points with straight-line interpolation with a mathematical modeling approach. A shape representation unit creates a mathematical model (such as a curved surface model or B-spline model) that represents the target object's shape. This mathematical model automatically generates smooth curved paths without requiring many training points, thus substituting the mechanical interpolation method with a mathematical modeling method that achieves higher accuracy with fewer points.
Solution Approach 2:
The patent changes the fundamental parameter of path generation from discrete point-based interpolation to continuous mathematical function-based modeling. By using mathematical models with adjustable parameters (such as control points for B-splines or surface parameters), the system can accurately represent complex curved shapes with minimal parameters, thereby reducing the number of training points needed while maintaining or improving path accuracy.
2Manufacturing precision
If the number of training points is increased to improve path accuracy along curved surfaces, then better approximation is achieved, but the training workload increases
Solution Approach 1:
The patent substitutes the time-consuming process of manually positioning and training at numerous discrete points with an automated mathematical modeling process. The shape representation unit automatically generates the mathematical model from a minimal set of training points, and the path generation unit automatically creates the action path from this model. This eliminates the manual workload associated with placing many training points while achieving the same or better accuracy.
Solution Approach 2:
The patent applies partial action by using only the minimum necessary training points to define the mathematical model, rather than using excessive numbers of points. The mathematical model then extrapolates the complete path information from these partial inputs, achieving full path accuracy with only partial training data, thereby significantly reducing training workload.
3Manufacturing precision
If direct training is performed at corners and connection points between shapes, then accurate path generation at these critical locations is possible, but the training becomes difficult and more complex
Solution Approach 1:
The patent replaces the difficult manual training process at corners and connection points with automated mathematical modeling. The shape representation unit automatically identifies and models corners and connection points as part of the overall shape model, and the path generation unit automatically ensures accurate path following at these locations. This eliminates the need for operators to manually train at these difficult locations while maintaining high accuracy.
Solution Approach 2:
The patent performs preliminary action by pre-defining the mathematical model of the target object's shape, including corners and connection points, before path generation. The mathematical model预先 captures the geometric features at these critical locations, so that when the action path is generated from the model, accuracy at corners and connection points is automatically ensured without requiring separate difficult training operations at these locations.
Data Source
AI summary
A reception section 102 receives a classification of a shape of a target object for a case in which a robot is to perform an action along the target object. An instruction section 104 instructs a user with a number of training points identified for each classification of the shape as training points for training positions of a specific location of the robot to perform the action. An acquisition section 106 acquires positional information of the specific location of the robot as trained by manipulation by the user according to the instruction. A estimation section 108 estimates a shape of the target object based on the acquired positional information. A generation section 110 generates a path for the robot to perform the action along the target object based on the estimated shape of the target object.