Mobile Robot Neural Network Path Learning for Dynamic Obstacles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mobile robots face inefficiencies in path planning due to the need to analytically avoid both stationary and moving obstacles, which increases calculation time and limits mobility, as existing methods do not effectively incorporate information on dynamic obstacles in map information.
Innovation Solution
A two-stage machine learning method for neural networks that first learns to avoid stationary obstacles through simulation and then adapts to avoid moving obstacles in real-time, using user input and virtual obstacle generation to enhance training data and reduce calculation complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the mobile robot analytically plans both travel path and avoidance path, then the robot can avoid obstacles, but the calculation time increases and mobility is limited
Solution Approach 1:
The patent segments the learning process into two distinct phases: first learning stationary obstacles during simulation, and second learning moving obstacles after deployment. This segmentation allows the robot to handle different obstacle types efficiently, reducing real-time calculation burden while maintaining comprehensive avoidance capability.
Solution Approach 2:
The patent applies preliminary action by pre-learning the patterns of stationary obstacles during the simulation phase before the robot is deployed. This advance learning stores knowledge about stationary obstacle avoidance, so that during real operation, the robot only needs to learn and react to moving obstacles, significantly reducing real-time calculation requirements.
2Productivity
If a neural network is used to avoid analytic calculations, then calculation speed improves, but it is difficult to make the neural network learn all combinations of stationary and moving obstacles
Solution Approach 1:
The patent divides the complex learning task into two manageable segments: stationary obstacle learning in simulation and moving obstacle learning in deployment. This segmentation transforms an intractable problem of learning all obstacle combinations into two simpler, sequential learning tasks that are computationally feasible.
Solution Approach 2:
The patent performs preliminary learning of stationary obstacles during the simulation phase, establishing a foundation of knowledge before deployment. This preliminary action reduces the complexity of real-time learning by pre-processing the more predictable stationary obstacle patterns, allowing the neural network to focus only on adapting to moving obstacles afterward.
3Adaptability or versatility
If the neural network learns to avoid all obstacles in real-time, then the robot adapts to dynamic environments, but the learning time and computational resources increase
Solution Approach 1:
The patent segments adaptability development into two stages: general spatial reasoning from stationary obstacles, and specific dynamic adaptation from moving obstacles. This segmentation allows the robot to achieve comprehensive adaptability more efficiently by building knowledge progressively rather than learning everything simultaneously.
Solution Approach 2:
The patent performs preliminary adaptation to obstacle avoidance patterns during simulation with stationary obstacles. This preliminary adaptation establishes baseline navigation skills, so that when deployed, the robot only needs to adapt to the variability of moving obstacles, reducing total learning time while maintaining high adaptability.
Data Source
AI summary
A machine learning method includes: a first learning step which is performed in a phase before a neural network is installed in a mobile robot and in which a stationary first obstacle is placed in a set space and the first obstacle is placed at different positions using simulation so that the neural network repeatedly learns a path from a starting point to the destination which avoids the first obstacle; and a second learning step which is performed in a phase after the neural network is installed in the mobile robot and in which, when the mobile robot recognizes a second obstacle that operates around the mobile robot in a space where the mobile robot moves, the neural network repeatedly learns a path to the destination which avoids the second obstacle every time the mobile robot recognizes the second obstacle.