Predictive Obstacle Maps for Low-Sensor Robot Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous robots rely on expensive high-frequency sensors for navigation, making them costly and limiting their widespread adoption due to difficulties in controlling movement near obstacles.
Innovation Solution
The use of predictive map generation technology based on LSTM neural networks, which generate sequences of historical maps to create predictive maps, allowing for unsupervised learning and anticipation of obstacle locations, thereby enhancing navigation and reducing the complexity of sensor usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If high-frequency sensors are used to ensure safety during navigation, then navigation reliability is improved, but system cost increases
Solution Approach 1:
The system performs preliminary mapping and obstacle prediction before navigation occurs. Historical maps are generated and stored in advance, and the LSTM network predicts future obstacle locations proactively, allowing the robot to plan paths ahead rather than reacting to immediate sensor data, thereby reducing reliance on expensive high-frequency sensors
Solution Approach 2:
The system creates simplified copies of the environment through historical maps and predictive models. Instead of relying on continuous real-time sensor data, the LSTM network generates predicted obstacle positions based on historical patterns, effectively copying essential spatial information in a computationally efficient manner that reduces sensor requirements
2Reliability
If high-frequency sensors are deployed to control robot movement near obstacles, then collision avoidance is improved, but device complexity increases
Solution Approach 1:
The patent replaces the mechanical sensor-based detection system with an information-processing approach using LSTM neural networks. Instead of relying on continuous mechanical sensor feedback to detect obstacles, the system uses learned patterns from historical maps to predict obstacle positions, substituting complex sensor hardware with a computational model that processes spatial-temporal data
3Reliability
If conventional navigation approaches are used to maintain safety distance from obstacles, then safety is improved, but ease of operation deteriorates due to difficulty in controlling movement
Solution Approach 1:
The system dynamically adjusts navigation behavior based on predicted obstacle positions rather than maintaining fixed safety distances. The LSTM network provides time-varying predictions of obstacle locations, allowing the robot to optimize its path dynamically - approaching obstacles when safe and maintaining distance when predicted collision risk exists, thereby improving both safety and movement control flexibility
Data Source
AI summary
Systems, apparatuses and methods may provide for technology that generates a sequence of predictive maps based on a sequence of historical maps and overlays the sequence of predictive maps on one another to obtain a map overlay. The technology may also apply an attenuation factor to the map overlay. In one example, the map overlay includes a grid of cells and each cell includes an occupation probability in accordance with the attenuation factor.