LSTM Persistence Filter for Digital Map Update Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital map data updating systems face challenges in accurately and efficiently processing sensor data from multiple vehicles to detect changes in geographic features, such as road signs, due to noise sensitivity and resource constraints, which can degrade the accuracy and safety of autonomous driving systems.
Innovation Solution
A machine learning-based persistence filter using a Recurrent Neural Network (RNN), specifically Long Short-Term Memory (LSTM) networks, processes sensor data to predict the presence or absence of geographic features, reducing noise sensitivity and the need for prior probability settings, allowing for real-time updates of digital map data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sensor data processing methods are used to update digital map data, then the system can detect changes in geographic features, but the system suffers from noise sensitivity and high resource requirements that degrade accuracy
Solution Approach 1:
The system performs preliminary actions by collecting and storing sensor observations over time before making update decisions. The persistence filter maintains a history of observations (positive and negative) for each geographic feature, allowing the system to evaluate multiple data points before determining whether a change has actually occurred. This preliminary data accumulation reduces noise sensitivity by requiring consistent evidence across multiple observations before triggering a map update.
Solution Approach 2:
The patent introduces a persistence filter as an intermediary component between raw sensor data and the digital map database. This intermediary layer processes sensor observations through machine learning models (LSTM/RNN) to determine whether observed changes represent true geographic feature changes or noise. The persistence filter acts as a buffer that filters out spurious detections while preserving genuine changes, thereby reducing noise sensitivity without losing detection accuracy.
2Reliability
If frequent updates of digital map data are performed to maintain accuracy, then the map data remains current, but the computational resources and processing time increase significantly
Solution Approach 1:
The system implements self-service through automated persistence filtering and change detection algorithms that operate autonomously on incoming sensor data. The machine learning models automatically evaluate observations against historical data and determine update necessity without human intervention. This automation enables frequent, reliable map updates while optimizing resource usage by only processing and storing data when actual changes are detected, rather than performing unnecessary update operations.
Solution Approach 2:
The patent employs parameter changes by using machine learning models (LSTM/RNN) that dynamically adjust their processing based on input patterns. The persistence filter learns from historical observation patterns and adapts its sensitivity thresholds, allowing it to maintain high detection accuracy while reducing computational overhead for routine, non-changes. The system changes its processing parameters based on the confidence level of detections and the stability of observed features.
3Measurement precision
If complex machine learning models are used to filter noise, then noise sensitivity is reduced, but the device complexity and resource requirements increase
Solution Approach 1:
The patent extracts the complex computational workload from the vehicle-based sensor systems and relocates it to centralized cloud-based processing servers. Individual vehicles collect and transmit raw sensor observations, but the actual machine learning inference and persistence filtering are performed remotely. This extraction allows the vehicle systems to remain relatively simple while still benefiting from sophisticated noise filtering and change detection capabilities provided by the cloud-based persistence filter.
Data Source
AI summary
An approach is provided for map updates using a machine learning-based persistence filter. The approach involves, for example, processing sensor data collected from vehicles to determine positive observations and/or negative observations of a map or geographic feature. The method also comprises providing the positive and/or observations as an input sequence to a machine learning model. The machine learning model, for instance, includes an input layer that feeds the input sequence to a long short-term memory (LSTM) layer or other type of recurrent neural network (RNN) layer. The LSTM or RNN layer connects to a fully connected layer and then to an output layer. The output layer outputs a predicted sequence of positive and/or negative observations. The method further comprises initiating an update of a geographic database to add, remove, or update the geographic feature based on the predicted sequence.