Redundant Perception Tracking for Automated Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated driving systems face challenges in accurately determining object positions and movements due to systematic errors in sensors, which can lead to incorrect vehicle maneuvers and potential collisions, as existing methods struggle to detect and account for errors in sensor data from overlapping fields of view.
Innovation Solution
The system generates multiple hypotheses about an object by excluding data from specific sensor types, allowing it to determine an object's state without relying on erroneous measurements, and performs vehicular maneuvers based on these hypotheses, thereby mitigating the impact of sensor errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If sensor data from multiple sensors with overlapping fields of view is used to detect objects, then measurement coverage is improved, but systematic errors in sensors can lead to incorrect object position determination
Solution Approach 1:
The patent segments the sensor data processing by creating multiple hypotheses, where each hypothesis excludes data from a specific sensor type. This allows the system to evaluate object positions independently of any single sensor's systematic errors, thereby maintaining measurement coverage while improving position determination accuracy through selective data exclusion in different hypothesis scenarios.
Solution Approach 2:
The patent changes the parameter of data inclusion by dynamically excluding environmental information from specific sensor types when generating different hypotheses. This parameter change allows the system to adapt to systematic errors by adjusting which sensor data is included in each hypothesis, resolving the contradiction between comprehensive coverage and accurate position determination.
2Reliability
If environmental information from all sensors is used to generate hypotheses, then tracking robustness is improved, but the system cannot detect or account for systematic errors in individual sensors
Solution Approach 1:
The patent extracts environmental information from specific sensor types when generating certain hypotheses, creating a set of hypotheses that deliberately exclude data from sensors that may contain systematic errors. This extraction approach allows the system to maintain tracking robustness through multiple hypotheses while simultaneously enabling error detection by comparing results across hypotheses with and without suspicious sensor data.
Solution Approach 2:
The patent implements feedback by comparing object position determinations across multiple hypotheses. When hypotheses that exclude specific sensor data produce significantly different results, the system can identify potential systematic errors in those sensors. This feedback mechanism maintains robustness through hypothesis diversity while enabling error detection through result comparison.
3Reliability
If multiple hypotheses excluding specific sensor data are generated, then error mitigation is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial action by generating multiple hypotheses that exclude data from specific sensor types rather than processing all possible sensor combinations. This selective approach provides sufficient error mitigation through targeted hypothesis exclusion without requiring exhaustive computational analysis of all sensor data permutations, thereby reducing overall computational complexity while maintaining reliability.
Data Source
AI summary
Redundant environment perception tracking for automated driving systems. One example embodiment provides a surveillance system, the system including a plurality of sensors, a memory, and an electronic processor. The electronic processor is configured to receive, from the plurality of sensors, environmental information of a common field of view, generate, based on the environmental information, a plurality of hypotheses regarding an object within the common field of view, the plurality of hypotheses including at least one set of hypotheses excluding the environmental information from at least one sensor of a first sensor type, determine, based on a subset of the plurality of hypotheses, an object state of the object, wherein the subset includes the at least one set of hypotheses excluding the environmental information from the at least one sensor of the first sensor type, and track the object based on the object state that is determined.