Runway Incursion Tracking With Probabilistic Trajectory Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems lack the ability to integrate low probability detections over time to ensure the probability requirements for missing a runway incursion (false negative) or reporting a non-existent incursion (false positive) are met, and they fail to predict future object trajectories accurately, especially in dynamic environments.
Innovation Solution
A system that integrates object detections over time, estimates object states, predicts future motion using Kalman filters and Gaussian processes, and filters detections based on pre-selected thresholds to determine the probability of runway incursions, adapting to new environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing end-to-end machine learning models are used for runway incursion detection, then the system can process detections quickly, but the system cannot build probability guarantees into the model output
Solution Approach 1:
The system segments the detection process into distinct modules: an end-to-end machine learning model for rapid object detection, a tracking module for associating detections across frames, and a probabilistic decision-making module for incursion determination. This segmentation allows each module to specialize - the ML model maintains high speed while the probabilistic module ensures reliability through formal probability guarantees.
Solution Approach 2:
The tracking module serves as an intermediary between the fast ML detector and the reliability-critical decision-making system. It maintains object trajectories and state estimates over time, enabling the system to integrate low probability detections across multiple frames while preserving both the speed of the original detector and the ability to provide probability guarantees in the final decision.
2Reliability
If the system integrates low probability detections over time to meet probability requirements, then the reliability improves, but the system complexity increases
Solution Approach 1:
The system employs dynamic tracking that adapts to varying detection probabilities and object behaviors. The tracker maintains probabilistic state estimates that evolve over time based on new detections, allowing the system to naturally integrate low probability detections when they are consistent with tracked trajectories while keeping the implementation manageable through established tracking algorithms.
Solution Approach 2:
The system uses feedback from multiple detection frames to update object state estimates and trajectory predictions. By continuously comparing new detections with predicted states and adjusting probability assessments accordingly, the system achieves reliable incursion determination through cumulative evidence while maintaining a structured approach that manages complexity.
3Measurement precision
If the system predicts future object trajectories to distinguish incursion types, then the accuracy of incursion classification improves, but the computational requirements increase
Solution Approach 1:
The system performs preliminary trajectory prediction using the tracked object state and motion model before making incursion classification decisions. By estimating future positions based on current state and velocity, the system can determine whether objects will clear the runway before landing without requiring full-blown simulation, thus achieving accurate classification with reduced computational energy.
4Adaptability or versatility
If the system adapts to new environments by detecting new object tracks, then the adaptability improves, but the false positive rate may increase
Solution Approach 1:
The system applies a threshold-based approach where object tracks are only declared as incursions when the accumulated probability evidence exceeds a predetermined threshold. This partial action principle allows the system to adapt to new environments by detecting novel object tracks while maintaining reliability by requiring sufficient evidence before declaring false positives, thus balancing adaptability with controlled false positive rates.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system and method for achieving a high confidence threshold when asserting that the runway is clear of obstacles. A landing aircraft periodically scans runway regions to confirm, track, and propagate trajectories of self-reporting and non-cooperative objects. An incurring object, appearing as a very tiny artifact, is unlikely to be detected in every sensor frame. The system and method propagate the trajectories of the incurring objects to confirm that the objects will clear the runway before the aircraft lands by tracking the objects and using a probabilistic decision maker.