Probabilistic Map Matching Using Hidden Markov Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing map matching methods are prone to errors due to inaccurate GPS and digital map data, particularly in urban areas where multiple streets are close together, leading to incorrect route identification and navigation.
Innovation Solution
The use of probabilistic models, such as Hidden Markov Models (HMMs), to determine the relative probability of routes based on factors like distance, time, velocity, historical information, and network complexity, allowing for the identification of the most likely route traversed by an entity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS location data is used for map matching, then real-time positioning is achieved, but accuracy deteriorates due to measurement noise and errors
Solution Approach 1:
The patent combines multiple candidate routes with different probabilities into a single probabilistic model. Instead of selecting just one route based on nearest distance, the system merges multiple possible routes and assigns probability values to each, allowing the most likely route to be identified while accounting for GPS measurement noise and urban complexity.
Solution Approach 2:
The patent introduces probability as a new parameter to evaluate routes. By transforming the deterministic nearest-route selection into a probabilistic framework, the system can handle measurement uncertainties and select routes based on likelihood rather than simple geometric proximity.
2Adaptability or versatility
If multiple candidate streets are considered in urban areas, then route options increase, but determination difficulty increases due to proximity of streets
Solution Approach 1:
The patent transforms the complex multi-route selection problem into a probabilistic evaluation. By assigning probability values to each candidate route based on multiple factors (distance, time, velocity, historical data), the system can efficiently determine the most likely route even when multiple streets are in close proximity.
Solution Approach 2:
The patent replaces the mechanical nearest-distance selection method with a probabilistic computation system. Instead of simply finding the closest road geometrically, the system uses probability calculations that incorporate multiple observational and contextual factors to identify the correct route.
3Productivity
If each noisy sensed location point is matched to the nearest path, then processing speed is maintained, but route accuracy deteriorates due to measurement noise
Solution Approach 1:
The patent combines multiple location points and their corresponding candidate routes into a unified probabilistic analysis. Instead of processing each point independently and potentially accumulating errors, the system merges information across multiple points to determine the overall most likely route sequence.
Solution Approach 2:
The patent replaces the simple nearest-path matching mechanism with a probabilistic evaluation system. This substitution allows the system to handle noisy measurements more effectively by considering the likelihood of each route given all observational data, rather than relying on direct geometric matching of individual noisy points.
Data Source
AI summary
Systems, methods, and devices are described for implementing map matching techniques relating to measured location data. Probabilistic models, including temporal Bayesian network models and Hidden Markov Models, may be used for combining multiple classes of evidence relating to potential locations of points traversed on routes over time. Multiple route segments and overall routes may be maintained under relative uncertainty as candidates. The candidate route segments and overall routes may then be reduced into a smaller number of candidates or a single most likely route as a trip progresses. As the trip progresses, route segments in proximity to each location point are identified and candidate matches are determined. A probability of an entity traversing a candidate match at a given time and a probability of an entity traversing between a first candidate match at a first time and a second candidate match at a second time are determined based on a plurality of factors. Different modalities may be used to measure and transmit the location data.