Predictive Map for Adaptive Sensor Fusion in Teleoperated Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing teleoperated vehicle control systems fail to provide an adequate representation of vehicle surroundings in adverse conditions such as bad weather or darkness, limiting operator control and safety.
Innovation Solution
A predictive map is used to dynamically fuse sensor data from various sources, including cameras, LIDAR, radar, and infrastructure sensors, to optimize the representation of vehicle surroundings based on location and situation, enabling proactive selection and weighting of sensors for improved visibility and safety.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If only camera sensor data is used for teleoperated vehicle control, then the system complexity is low, but the representation of vehicle surroundings is inadequate in adverse conditions such as bad weather or darkness
Solution Approach 1:
The system dynamically adapts the sensor fusion configuration based on detected environmental conditions. The predictive map is updated in real-time to reflect current weather, lighting, and location conditions, automatically adjusting which sensors are activated and how their data is weighted and fused. This dynamic adaptation ensures adequate surroundings representation in adverse conditions without permanently maintaining high system complexity.
Solution Approach 2:
The system pre-generates a predictive map that anticipates upcoming environmental conditions along the vehicle's route. By detecting current conditions and predicting future states, the system proactively configures sensor fusion parameters before adverse conditions occur, ensuring seamless transition to appropriate sensor configurations when needed without sudden complexity increases.
2Reliability
If sensor data from multiple sources is continuously fused, then the surroundings model is comprehensive, but the data processing and transmission load increases
Solution Approach 1:
The system applies different fusion strategies to different spatial regions and sensor types based on their specific characteristics and current relevance. Instead of uniformly processing all sensor data at full resolution, the predictive map identifies which local regions require enhanced monitoring and adjusts fusion intensity accordingly, reducing overall processing load while maintaining comprehensive coverage where critical.
Solution Approach 2:
The system dynamically changes processing parameters such as data sampling rates, fusion depth, and transmission resolution based on environmental conditions and vehicle context. When conditions are favorable, processing is reduced; when adverse conditions are detected or predicted, processing intensity is increased only for relevant sensors and regions, optimizing energy efficiency while maintaining model comprehensiveness.
3Adaptability or versatility
If the predictive map is updated in real-time with current environmental conditions, then the sensor selection and fusion is optimized, but the computational requirements increase
Solution Approach 1:
The predictive map system segments the environmental conditions into discrete categories (e.g., lighting conditions, weather states, road types) and pre-computes optimal sensor configurations for each segment. This segmentation allows the system to quickly lookup and apply appropriate configurations without performing complex real-time optimization, reducing computational requirements while maintaining high adaptability to changing conditions.
4Reliability
If more sensors are activated and their data is fused, then the visibility and safety are improved, but the device complexity and cost increase
Solution Approach 1:
The system designs the sensor fusion architecture to be universal and multi-functional, where the same fusion mechanisms and processing pipelines can handle data from different sensor types (camera, LIDAR, radar, infrastructure sensors). This universal framework reduces overall system complexity by avoiding separate processing paths for each sensor type, allowing comprehensive sensor activation without proportional increases in fusion mechanism complexity.
Data Source
AI summary
A method for the control of a vehicle by an operator. The method includes: using a predictive map to control the vehicle by: detecting a situation and/or location reference of the vehicle, transmitting data of a defined set of sensors, fusing and processing the data of the defined set of sensors; displaying the fused and processed data for the operator; creating/updating the predictive map by: recognizing a problematic situation and/or a problematic location by observation of the operator and/or marking by the operator, storing the problematic situation and/or the problematic location in a first database for storing problematic situations and locations, and training a model for selecting the defined set of sensors and fusing the data of the defined set of sensors by machine learning.