Autonomous Vehicle Map Priors for Intersection Creep Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in predicting the behavior of other vehicles on the road that deviate from conventional driving rules due to reasons like driver negligence or unforeseen circumstances, leading to potential safety hazards.
Innovation Solution
The integration of a probability layer into the vehicle's map database, using historical data to predict the behavior of other vehicles by analyzing systemic behaviors and adding this information as 'map priors', allowing the vehicle to anticipate and respond to potentially dangerous maneuvers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the autonomous vehicle follows map rules strictly, then navigation safety is improved, but the ability to predict and respond to erratic vehicle behavior deteriorates
Solution Approach 1:
The system performs preliminary actions by collecting historical vehicle behavior data and generating probabilistic predictions of future maneuvers before they occur. The map database is pre-populated with statistical information about likely vehicle actions at specific locations, enabling the AV to anticipate erratic behavior rather than merely react to it.
Solution Approach 2:
The system transitions from static map rules to dynamic probabilistic predictions. Instead of fixed navigation paths, the AV uses continuously updated probability distributions that adapt to observed vehicle behaviors, allowing the system to flexibly respond to changing traffic patterns and unpredictable driver actions.
2Measurement precision
If kinematic prediction models are used to observe and predict vehicle behavior, then future position prediction is improved, but accuracy for systemic maneuvers deteriorates
Solution Approach 1:
The system introduces map-based probabilistic information as an intermediary between raw kinematic observations and final behavior predictions. Instead of relying solely on direct observation of vehicle motion, the system queries pre-computed probability distributions from the map database that encode collective knowledge about vehicle behaviors at specific locations, thereby improving prediction reliability.
Solution Approach 2:
The system uses historical behavior data to create probabilistic models that copy and represent typical vehicle maneuvers at specific locations. These probabilistic copies of observed behaviors are stored in the map database and can be rapidly queried to predict future actions without requiring complex real-time analysis of each vehicle's kinematics.
3Ease of operation
If the map database includes only conventional driving rules, then navigation compliance is improved, but the ability to handle real-world erratic behavior deteriorates
Solution Approach 1:
The system merges conventional map rules with probabilistic behavior predictions into a unified navigation framework. The map database contains both explicit traffic rules and statistical information about actual vehicle behaviors, allowing the AV to combine rule-based navigation with data-driven predictions of other vehicles' actions.
Solution Approach 2:
The system adds a probabilistic dimension to traditional deterministic map data. Instead of storing only fixed road rules and geometry, the map database incorporates probability distributions for vehicle behaviors, transforming the navigation problem from a deterministic path-following task to a probabilistic decision-making process that accounts for uncertainty in other drivers' actions.
Data Source
AI summary
Systems, methods, and devices are disclosed for predicting creep behaviors of objects (vehicles, bicycles, pedestrians, etc.) at a location. An autonomous vehicle located at a first road segment connecting to an intersection having a stop sign can detect a first vehicle approaching the intersection from a second road segment connecting to the intersection. Using a model indicating an average creep point location specific to the second road segment connecting to the intersection, the autonomous vehicle can predict that the first vehicle will yield at the average creep point location that is specific to the second road segment.