3D Surface Printing Path Planning With Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for printing images on complex 3D surfaces, especially curved surfaces like aircraft, face challenges in planning control paths for robotic printheads due to high degrees of freedom, leading to inefficiencies in printing time and quality.
Innovation Solution
A reinforcement learning-based system that generates simulated control paths for robotic printing systems using a processor with modules for simulated control path generation, reward value determination, and selection, incorporating 3D digital models and real-time sensor data to optimize printhead movement and ink deposition on contoured surfaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If robotic printing systems are used to print on complex 3D surfaces, then printing speed and productivity are improved, but control path planning becomes significantly more complex due to high degrees of freedom
Solution Approach 1:
The system performs preliminary 3D scanning of the contoured surface to create a digital model before printing. Control paths are pre-calculated and optimized based on this digital model, allowing the robotic system to execute printing operations without real-time complex calculations. This preliminary preparation resolves the contradiction by shifting computational complexity from the printing execution phase to the pre-processing phase.
Solution Approach 2:
The patent replaces traditional mechanical control path planning methods with machine learning-based algorithms. The neural network model learns optimal control paths from training data and predicts paths for new surfaces, substituting complex mechanical control logic with intelligent software-based solutions that handle high degrees of freedom more efficiently.
2Manufacturing precision
If printheads move at very close distance to the surface (less than 8 mm), then printing precision is improved, but the risk of collision with contoured surfaces increases
Solution Approach 1:
The system performs preliminary 3D scanning to create an accurate digital model of the contoured surface geometry. Control paths are pre-calculated to maintain optimal distance (less than 8 mm) from the surface while avoiding collision zones. This preliminary geometric analysis enables the system to achieve high printing precision without collision risk by knowing the exact surface topology in advance.
Solution Approach 2:
The system incorporates sensors that provide real-time feedback on the distance between printheads and the contoured surface. This feedback is fed back to the control system, which adjusts the control path dynamically to maintain optimal printing distance while preventing collisions. The closed-loop control resolves the contradiction by continuously monitoring and adjusting the system state.
3Manufacturing precision
If multiple printheads dispersing different color inks are used, then color accuracy is improved, but coordination and synchronization of printhead movement becomes more challenging
Solution Approach 1:
The patent merges the control of multiple printheads into a unified control framework. The machine learning model learns to coordinate all printheads simultaneously, treating them as an integrated system rather than separate entities. This merging approach reduces coordination complexity by establishing synchronized control patterns that maintain color accuracy while simplifying the overall control architecture.
Solution Approach 2:
The control system is designed with universal coordination algorithms that can manage any number of printheads with different ink colors. The same control framework and machine learning model handle coordination for all printheads regardless of their specific function or color output, reducing the need for separate coordination mechanisms for each printhead.
4Device complexity
If conventional printing methods are used on contoured surfaces, then equipment simplicity is maintained, but printing time and operational efficiency deteriorate
Solution Approach 1:
The patent replaces conventional mechanical printing methods with a machine learning-based control system. The neural network model predicts optimal control paths and printing parameters, substituting time-consuming trial-and-error or rule-based mechanical control with intelligent algorithms that compute solutions much faster, thereby improving printing efficiency while keeping the physical printing hardware relatively simple.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed herein is a livery printing system and a method of generating a control path. The system includes a training system having a processor and a memory with code configured to cause the processor to receive a 3D digital model associated with an object; generate simulated control paths, based on the 3D digital model, for actuators of a printing device with printheads, determine a reward value for each one of the simulated control paths based on a simulated physical value, a simulated surface coverage value, or a simulated printing speed value. A value of one simulated control path variable of any of the simulated control paths is different than the value of the simulated control path variable of another simulated control path. One of the simulated control paths is selected based on a comparison between the reward values determined for the simulated control paths.