Robot Navigation Using Distance Sensors and Neural Decision-Making
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional robot navigation in unfamiliar environments faces challenges due to complex input data and high uncertainty, leading to long delays and futile travel, particularly in the absence of a pre-created map.
Innovation Solution
A robot equipped with distance sensors and a neural network model that processes sensed distances and movement information to make real-time decision-making, enabling navigation without a pre-created map by iteratively refining movement decisions through a neural network model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional algorithms or reinforcement learning are used to explore walking directions in unfamiliar environments, then the robot can navigate without a pre-created map, but the complexity of input data and high uncertainty lead to long delays and futile travel
Solution Approach 1:
The patent transforms the navigation problem by changing the parameter representation from complex raw sensor data to simplified relative distance relationships. The neural network model processes relative distances between the robot and surrounding objects, converting high-dimensional uncertain data into meaningful spatial relationships that guide navigation decisions, thereby reducing exploration time while maintaining adaptability
Solution Approach 2:
The patent introduces relative distance as an intermediary parameter between the robot's position and surrounding objects. Instead of directly processing complex sensor data, the system uses relative distance measurements to create a simplified representation of the environment, which serves as a mediator for the neural network to make navigation decisions, reducing both data complexity and uncertainty
2Adaptability or versatility
If Lidar or depth camera is used to check relative distance and explore walking directions, then the robot can operate without a map, but the input data complexity increases and training becomes difficult
Solution Approach 1:
The patent extracts only the essential information needed for navigation from the complex sensor data. Instead of using all available data from Lidar or depth cameras, the system selectively extracts relative distance information between the robot and surrounding objects, discarding redundant data and reducing input complexity while maintaining the ability to operate without a pre-created map
Solution Approach 2:
The patent changes the parameter representation from raw sensor data to relative distance measurements. By transforming the input data into relative distance relationships, the system reduces the dimensionality and complexity of the input while preserving the essential spatial information needed for navigation decisions in unfamiliar environments
Data Source
AI summary
A navigation method applicable to a robot includes: (a) setting a first position coordinate and first movement information; (b) measuring a plurality of to-be-sensed distances in different directions by using a plurality of distance sensors; (c) inputting the plurality of sensed distances, the first position coordinate, and the first movement information into a neural network model to obtain second movement information; (d) setting the second movement information as the first movement information for a next round of a decision-making process; (e) driving, based on the second movement information, the robot to move from the first position coordinate to a second position coordinate; (f) setting the second position coordinate as the first position coordinate for a next round of the decision-making process; and (g) repeating steps (b) to (f) until a distance between the second position coordinate and a destination coordinate is less than a threshold.