Sensor Data Fusion for Autonomous Vehicle Perception
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face limitations in perception due to the constraints of their immediate sensors and lines of sight, leading to inefficiencies in coordination among smart objects and potential safety issues in decision-making processes, especially when predicting the trajectories of other objects.
Innovation Solution
A method that involves receiving and merging sensor data from multiple sources to generate an augmented dataset, which is then transmitted to the vehicle, enhancing its perception and decision-making capabilities by providing more accurate and comprehensive location information about surrounding objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data is collected only from the autonomous vehicle's own sensors, then the vehicle maintains operational independence and quick decision-making, but the perception accuracy and awareness of surroundings remain limited to immediate vicinity and lines of sight
Solution Approach 1:
A computing apparatus acts as an intermediary between multiple sensing objects (autonomous vehicles and stationary sensing objects) to collect, merge, and process sensor data. This intermediary consolidates perception information from diverse sources and transmits augmented datasets to target vehicles, enabling enhanced perception without requiring direct peer-to-peer communication between all sensing objects.
Solution Approach 2:
The system merges sensor data from multiple sources including autonomous vehicles' own sensors and stationary sensing objects' sensors. The computing apparatus combines these datasets to create augmented perception data that provides broader environmental awareness beyond what any single vehicle could perceive alone, overcoming limitations of immediate vicinity and lines of sight.
2Loss of information
If sensor data from multiple smart objects is collected and merged, then the completeness and accuracy of environmental perception is improved, but the coordination efficiency and response time may be reduced due to data processing overhead
Solution Approach 1:
Stationary sensing objects continuously collect and pre-process sensor data in advance, maintaining updated perception datasets ready for transmission. When an autonomous vehicle requests environmental information, the computing apparatus can quickly retrieve and transmit pre-augmented data rather than collecting and processing data in real-time, reducing coordination delay.
Solution Approach 2:
The computing apparatus transmits selective excerpts of augmented datasets tailored to each autonomous vehicle's specific needs, location, and sensing capabilities. Rather than transmitting complete datasets from all sources, only relevant information is shared, reducing data transmission and processing time while maintaining information completeness for decision-making.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Embodiments include a method comprising, at a computing apparatus: receiving first sensor data from one or more sensors provided by a first sensing object, the first sensing object being an autonomous or partially autonomous mobile object, the mobile object having wireless connectivity, the first sensor data representing a state of an entity in the physical domain; receiving second sensor data from one or more sensors provided by a second sensing object, the second sensing object being distinct from the first sensing object, the second sensor data representing the state of the entity in the physical domain; merging the first sensor data and the second sensor data to generate an augmented dataset representing the state of the entity; transmitting an excerpt of the augmented dataset to the mobile object, the excerpt being a selected subset of the augmented dataset, the selected subset being selected by the computing apparatus in dependence upon the identity or location of the mobile object.