Sensor Fusion Obstacle Detection Using Bayesian Probability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle safety systems face challenges with noisy, sparse, and inaccurate sensor data, leading to false positives and unnecessary interventions due to 'ghost objects', which can cause inconvenience and potential harm.
Innovation Solution
A sensor fusion system that combines data from imaging devices and sensors using a collision avoidance module, incorporating a data association module, tracking module, probability module, obstacle identification, collision prediction, and decision-making to accurately track and predict obstacles, reducing false positives through Kalman filtering and Bayesian probability updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor data is used for obstacle detection, then obstacle detection capability is provided, but false positives and ghost objects occur due to noisy and sparse data
Solution Approach 1:
The patent combines data from multiple sensors (RADAR, LIDAR, cameras, ultrasonic sensors) to create a comprehensive view of the environment. By fusing data from these diverse sources, the system cross-validates detections and eliminates false positives that would occur with any single sensor, directly resolving the contradiction between providing obstacle detection capability and avoiding false positives from noisy data
Solution Approach 2:
The patent introduces a probabilistic framework as an intermediary layer between raw sensor data and obstacle detection decisions. This framework uses probability thresholds and confidence levels to filter out ghost objects and noisy detections, allowing the system to maintain high detection capability while achieving reliable accuracy by mediating between imperfect sensor data and detection outcomes
2Object-affected harmful factors
If active safety features intervene on detected objects, then collision avoidance is achieved, but unnecessary interventions occur due to false positives
Solution Approach 1:
The patent applies preliminary filtering and validation steps before triggering active safety interventions. By pre-processing sensor data through probabilistic frameworks and cross-sensor validation, the system eliminates false positives before they reach the decision-making stage, preventing unnecessary interventions while maintaining protection against real collisions
Solution Approach 2:
The patent implements feedback loops where detection results and intervention outcomes are continuously monitored and used to refine future detections. This feedback mechanism learns from false positives and adjusts detection thresholds, reducing unnecessary interventions over time while maintaining collision avoidance capability
Data Source
AI summary
A controller receives outputs form a plurality of sensors such as a camera, LIDAR sensor, RADAR sensor, and ultrasound sensor. Sensor outputs corresponding to an object are assigned to a tracklet. Subsequent outputs by any of the sensors corresponding to that object are also assigned to the tracklet. A trajectory of the object is calculated from the sensor outputs assigned to the tracklet, such as by means of Kalman filtering. For each sensor output assigned to the tracklet, a probability is updated, such as using a Bayesian probability update. When the probability meets a threshold condition, the object is determined to be present and an alert is generated or autonomous obstacle avoidance is performed with respect to an expected location of the object.