Recursive Filter for Collision Detection Impact Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing collision detection systems face inaccuracies in predicting impact time and velocity due to sensor measurement errors caused by the geometry of target objects, especially when short-range sensors cannot impinge on the nearest point of the target, leading to biased readings and non-Gaussian errors.
Innovation Solution
A system that utilizes a recursive filter, such as a Kalman filter, to refine impact time and velocity predictions by disregarding sensor readings when the target object is within a certain threshold distance, preventing the inclusion of inaccurate data and relying on previous accepted readings for calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor readings are continuously used to refine impact predictions, then the system can maintain updated estimates, but measurement errors from sensor geometry limitations cause prediction inaccuracies
Solution Approach 1:
The patent extracts and removes problematic sensor readings from the prediction process when the sensor cannot impinge on the nearest point of the target. By identifying readings that fail to meet geometric criteria (where the sensor beam does not strike the closest point on the target object), the system excludes these erroneous measurements from refining impact predictions, thereby preventing them from degrading prediction accuracy.
Solution Approach 2:
The patent applies different quality standards to different sensor readings based on local geometric conditions. Rather than uniformly accepting or rejecting all readings, the system evaluates each reading's geometric validity - specifically whether the sensor beam impinges on the nearest point of the target - and selectively processes only those readings that meet the local quality criterion of geometric correctness.
2Adaptability or versatility
If short-range sensors are mounted at high positions with downward angles to cover broad vertical spectrum, then detection coverage is improved, but the sensor may not impinge on the nearest point of certain target geometries causing biased readings
Solution Approach 1:
The patent introduces dynamic evaluation of sensor reading validity based on real-time geometric conditions. The system continuously assesses whether each sensor reading meets the criterion of impinging on the nearest point of the target, adapting the acceptance criteria for readings based on the current relative geometry between sensor and target rather than using fixed geometric assumptions.
Solution Approach 2:
The patent changes the parameter of reading acceptance from a fixed geometric assumption to a dynamic validity check. By introducing a geometric validity parameter that evaluates whether the sensor beam strikes the nearest point on the target, the system transforms the rigid mounting geometry into a flexible validation criterion that can accommodate various target shapes and positions.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly improves the accuracy of impact time and velocity predictions, reducing errors and ensuring more reliable activation of safety features in vehicles, with demonstrated improvements in prediction accuracy up to 90% compared to prior methods.
Implementation Method 1
A LIDAR system uses lasers to provide an accurate, high-frequency signal identifying a close-range object
Implementation Method 2
beams from the short range sensors are likely to contact a target at the point closest to the host vehicle, providing accurate distancing information to the sensors
Data Source
AI summary
A collision detection system includes at least one sensor. The data from the sensor is input to a recursive filter which selectively uses the input to predict the motion of a target object relative to a host vehicle. The recursive filter continues to use the data from the sensor until the target object is within a threshold distance of the host vehicle. Within the threshold distance, the sensor does not reliably provide accurate data. Accordingly, the recursive filter omits the sensor input from motion estimates when the target object is within the threshold distance, which leads to significant improvement in target motion prediction.