Freespace Detection Using Machine Learning for Autonomous Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous machines face challenges in navigating complex environments due to the high cost and computational complexity of high-precision sensors, which can be limited by inclement weather and lighting conditions, and often require multiple sensors to achieve accurate freespace detection, increasing operational expenses and reducing accessibility.
Innovation Solution
A machine learning model is trained using data from a combination of sensors, such as LiDAR and RADAR, where freespace labels from high-resolution LiDAR data are propagated to RADAR data, enabling the model to identify navigable areas using less expensive and computationally less intensive RADAR data, improving navigation accuracy and reducing sensor requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high precision sensors (LiDAR) are used to detect freespace, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent introduces an intermediary approach by using a lower-cost RADAR sensor as a mediator to capture freespace data, which is then enhanced through machine learning models trained on LiDAR data. This allows the system to achieve high measurement precision without directly using expensive LiDAR sensors in all applications, thereby reducing device complexity while maintaining accuracy.
Solution Approach 2:
The patent creates a copy of the LiDAR data processing pipeline by training machine learning models on LiDAR-generated freespace labels and then applying these models to RADAR data. This copying approach enables the system to replicate the precision of LiDAR-based detection using more affordable RADAR sensors, thus reducing overall device complexity and cost.
2Measurement precision
If high precision sensors (LiDAR) are used to detect freespace, then measurement precision is improved, but cost increases
Solution Approach 1:
The patent employs cheaper RADAR sensors instead of expensive LiDAR sensors for actual deployment, using the trained machine learning models to compensate for the lower cost. This principle allows the system to achieve acceptable measurement precision while significantly reducing the cost of sensor hardware, making autonomous navigation more accessible.
Solution Approach 2:
By copying the LiDAR processing pipeline through machine learning models trained on LiDAR data, the system can use inexpensive RADAR sensors to achieve similar detection precision. This copying strategy reduces the overall system cost while maintaining the ability to accurately detect freespace for navigation purposes.
3Measurement precision
If multiple sensors are used to achieve accurate freespace detection, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts the essential freespace detection function from complex multi-sensor systems and implements it using a single RADAR sensor enhanced by machine learning models. By taking out the core detection capability and separating it from the complexity of multiple sensors, the system achieves accurate freespace detection with reduced device complexity.
Solution Approach 2:
The machine learning model serves as an intermediary that processes RADAR data and produces freespace detection results similar to what would be achieved with multiple high-precision sensors. This intermediary approach maintains measurement precision while significantly reducing the number of physical sensors required, thereby lowering device complexity.
Data Source
AI summary
Systems and methods are disclosed that relate to freespace detection using machine learning models. First data that may include object labels may be obtained from a first sensor and freespace may be identified using the first data and the object labels. The first data may be annotated to include freespace labels that correspond to freespace within an operational environment. Freespace annotated data may be generated by combining the one or more freespace labels with second data obtained from a second sensor, with the freespace annotated data corresponding to a viewable area in the operational environment. The viewable area may be determined by tracing one or more rays from the second sensor within the field of view of the second sensor relative to the first data. The freespace annotated data may be input into a machine learning model to train the machine learning model to detect freespace using the second data.