Distance Object Detection Training Using Aligned Multi-Vehicle Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles equipped with sensors face challenges in accurately detecting objects at a distance due to less precise data acquisition, leading to potential false negatives and inadequate corrective actions.
Innovation Solution
A method where a second vehicle follows a first vehicle equipped with similar sensors, allowing both to acquire data simultaneously, which is then aligned and used to train a machine learning algorithm to improve object detection in less accurate representations, enhancing the algorithm's ability to recognize objects in varying physical settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Length of stationary object
If sensor data is acquired from a second vehicle at a greater distance, then the detection range is extended, but the measurement precision deteriorates
Solution Approach 1:
The patent combines sensor data from two vehicles - a first vehicle at a closer distance and a second vehicle at a greater distance. The server receives and merges the sensor data from both vehicles, allowing the system to extend detection range while maintaining precision through data fusion. The first vehicle's high-precision data compensates for the second vehicle's lower precision, resolving the contradiction between range and accuracy.
Solution Approach 2:
The server acts as an intermediary that receives, processes, and aligns sensor data from both vehicles. It synchronizes the data in time and space, then uses the first vehicle's accurate data to train machine learning algorithms that can interpret the second vehicle's less precise but更远 range data, mediating between the two quality levels.
2Measurement precision
If machine learning algorithm is trained on high-precision data only, then the recognition accuracy is improved, but the adaptability to varying physical settings deteriorates
Solution Approach 1:
The patent changes the training parameters of the machine learning algorithm by incorporating data from multiple distance settings. Instead of training only on high-precision close-range data, the algorithm is trained on a mixture of data from the first vehicle (close, high precision) and second vehicle (far, lower precision). This parameter change in training data diversity improves the algorithm's adaptability to various detection scenarios while maintaining reasonable accuracy.
3Device complexity
If the system uses data from a single vehicle, then the device complexity is reduced, but the reliability of object detection deteriorates
Solution Approach 1:
The system merges sensor data from two vehicles to improve detection reliability. The server combines the data streams, allowing cross-validation and compensation for individual sensor limitations. This data merging approach reduces false negatives and improves overall detection reliability without requiring each individual vehicle to have complex redundant sensor systems.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach improves the accuracy of object detection in autonomous vehicles by aligning and re-training the machine learning algorithm to handle less precise sensor data, enabling more effective recognition and response to potential hazards.
Implementation Method 1
a first LIDAR sensor and a second LIDAR sensor mounted respectively on a first vehicle and a second vehicle
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There is provided a method and a server for training a machine-learning algorithm (MLA) to detect objects in sensor data acquired by a second sensor located at a second physical setting from the objects, the MLA having been trained to recognize the objects in sensor data acquired by a first sensor located at a first physical setting from the objects. First sensor data acquired by the first sensor on a first vehicle is aligned spatially and temporally with second sensor data acquired by the second sensor on a second vehicle. Objects are detected in portions of the aligned first sensor data and classified by the MLA. The object classes in the aligned first sensor data are assigned to corresponding portions in the aligned second sensor data. The MLA is trained on the labelled portions in the aligned second sensor data to recognize and classify objects at the second distance.