Convolutional Neural Network Heuristics Map for Path Searching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing path searching technologies face challenges in versatility and efficiency, particularly in acquiring optimal heuristics functions for complex tasks and requiring extensive manual setup, which limits their applicability and increases search time.
Innovation Solution
An information processing device and method utilizing a convolutional neural network (CNN) to learn a heuristics function based on obstacle maps, outputting a heuristics map as a two or more dimensional image, and employing this map for efficient path searching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional path searching algorithms are used, then path searching can be performed, but the search time is excessively long and the system lacks versatility
Solution Approach 1:
The patent replaces conventional mechanical search algorithms with a neural network-based system. A neural network is trained to predict optimal path sequences directly from environmental inputs (obstacle maps, start/end positions), eliminating the need for time-consuming traditional search algorithms like A* or Dijkstra. The neural network processes input maps and outputs complete path sequences in milliseconds, achieving both speed and adaptability across different environments.
Solution Approach 2:
The patent changes the fundamental parameters of path searching by using deep learning models instead of classical algorithms. The system transforms environmental data into feature representations through neural network layers, learning optimal paths through parameter optimization during training. This allows the system to adapt to various environments without manual reconfiguration, achieving both rapid search times and high versatility through learned parameters.
2Reliability
If manual setup and configuration are performed for path searching, then the system can be configured for specific tasks, but the process is time-consuming and reduces adaptability
Solution Approach 1:
The patent implements self-service through automated neural network training. The system automatically learns optimal path finding capabilities by training on diverse environmental data and task sequences. During operation, the trained model independently processes new environments and generates paths without requiring manual setup or configuration. This automated learning process eliminates time-consuming manual configuration while maintaining high reliability through learned best practices.
Solution Approach 2:
The patent applies preliminary action by pre-training the neural network on extensive datasets of environmental maps and optimal paths before actual use. This preliminary training phase allows the system to store learned patterns and heuristics in its neural parameters. When deployed to new tasks, the system retrieves and applies relevant learned patterns instantly without manual configuration, achieving both rapid response and task-specific adaptability through pre-acquired knowledge.
3Adaptability or versatility
If traditional search algorithms are used, then path searching is possible, but the system requires extensive manual setup and lacks adaptability to different environments
Solution Approach 1:
The patent achieves universality through a single neural network model that can handle multiple environmental types and task configurations. The network is trained to process various input map formats, obstacle configurations, and boundary conditions, enabling it to adapt to different environments without requiring separate specialized algorithms. This multi-functional approach eliminates the need for complex environment-specific setup procedures while maintaining high adaptability.
Data Source
AI summary
To enable learning of versatile heuristics with a large reduction in search time. Provided is an information processing device including: a learning unit configured to learn a heuristics function according to path searching, with a convolutional neural network, in which the convolutional neural network carries out learning based on a plurality of obstacle maps, to output a heuristics map expressing the heuristics function as a two or more dimensional image. Moreover, provided is an information processing method including: learning a heuristics function according to path searching, by a processor, with a convolutional neural network, in which the convolutional neural network carries out learning based on a plurality of obstacle maps, to output a heuristics map expressing the heuristics function as a two or more dimensional image.