Probability Distribution Map for Reliable Path Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for building obstacle maps fail to accurately represent the properties of objects, leading to unreliable and unstable path planning for robots and self-driving cars, as they do not account for kinematic, shape, and probabilistic properties of objects, resulting in potential getting stuck in local minima.
Innovation Solution
An apparatus and method that collect sensor information to recognize objects and create a probability distribution map by integrating object property models, including kinematic, shape, and probabilistic properties, along with system properties, to produce a reliable and stable map for optimal path planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional bottom-up methods are used to build obstacle maps by recognizing all objects only as obstacles, then the map building process is simple and fast, but the reliability and accuracy of the map is low because object properties are not considered
Solution Approach 1:
The patent segments objects into different categories (pedestrians, vehicles, two-wheeled vehicles, walls, etc.) based on their properties, and applies different probability distribution models to each category. This segmentation allows the system to consider specific object properties while maintaining a structured approach to map building, resolving the contradiction between reliability and complexity.
Solution Approach 2:
The patent changes the parameters used to represent objects from uniform obstacle representation to property-specific parameters (kinematic properties, shape properties, probabilistic properties). By adjusting these parameters according to object type, the system achieves higher map reliability without excessive complexity through systematic parameter management.
2Measurement precision
If object properties such as kinematic, shape, and probabilistic properties are integrated into the map, then the path planning accuracy and reliability improve, but the computation time and processing complexity increase
Solution Approach 1:
The patent applies probability distribution models to objects in advance based on their recognized properties, creating pre-computed probability maps. This preliminary action allows the system to have path planning information ready before actual navigation, reducing real-time computation time while maintaining high precision through pre-analyzed object properties.
Solution Approach 2:
The patent applies different probability distribution models locally to different object types rather than using a uniform approach. Each object category receives appropriate modeling (e.g., different kinematic models for vehicles vs. pedestrians), which improves path planning precision for each specific case while avoiding the overhead of overly complex global models.
3Reliability
If conventional methods represent all obstacles with the same property, then the map building process is simple, but the map has low reliability and cannot accurately calculate arrival times at obstacles
Solution Approach 1:
The patent introduces dynamic probability distribution models that adapt to different object properties and movements. Instead of static uniform representation, the system uses dynamic models that update probability distributions based on object characteristics and motion, improving safety map reliability while managing detection complexity through standardized dynamic modeling approaches.
Solution Approach 2:
The patent creates simplified probability distribution representations (copies) of complex object properties. Rather than directly processing all detailed object characteristics, the system generates probabilistic models that capture essential features needed for safety assessment, making property detection more manageable while maintaining high reliability in the resulting maps.
Data Source
AI summary
An apparatus and method for building a map of probability distribution are provided. The apparatus for building the map of probability distribution includes: a sensor information collector configured to collect sensor information from a plurality of sensors; as object recognizer configured to recognize an object by integrating and inferring the sensor information, and to acquire object information; and a probability distribution creator configured to determine whether to apply an object property model including at least one of kinematic properties, shape properties, and probabilistic properties in correspondence to the object information, to acquire object properties corresponding to the object information, and to create a probability distribution based on foe object properties. Accordingly, it is possible to build a map with high reliability.