Autonomous Vehicle Sensor Fusion Against Jamming Attacks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles are vulnerable to jamming attacks that disrupt sensors, leading to potential accidents and safety threats, as demonstrated by successful experiments where LiDAR systems were jammed, causing objects to disappear, resulting in dangerous situations.
Innovation Solution
A computer-implemented method that utilizes multiple vehicle sensors to detect the presence of an object after it is no longer detected by one sensor, triggering safety measures such as alerts and manual control switches, and employs machine learning and AI to predict object trajectories and mitigate jamming attacks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single sensor is used to detect objects for autonomous vehicle navigation, then the device complexity is reduced, but the reliability of object detection deteriorates when subjected to jamming attacks
Solution Approach 1:
The system segments the detection task across multiple sensor types (LiDAR, radar, cameras, ultrasonic sensors) rather than relying on a single sensor. Each sensor independently monitors for objects, and the system integrates their outputs to maintain reliable detection even when one sensor is jammed. This segmentation allows the system to tolerate sensor failures while maintaining overall detection reliability.
Solution Approach 2:
The system changes the operational parameters by switching between different sensor modalities based on detection needs. When LiDAR is jammed, the system transitions to using radar or camera-based detection, effectively changing the detection parameter from optical range-finding to electromagnetic wave reflection or image processing, thereby maintaining detection reliability under adversarial conditions.
2Reliability
If multiple sensors are deployed to detect objects and mitigate jamming attacks, then the reliability of object detection is improved, but the device complexity increases
Solution Approach 1:
The system implements multi-functionality by using a unified object detection framework that processes inputs from multiple sensor types through a common neural network architecture. The same detection model handles data from LiDAR, radar, cameras, and ultrasonic sensors, allowing the system to maintain reliability across different sensor modalities without proportionally increasing processing complexity.
Solution Approach 2:
The system creates virtual copies of detection capabilities across different sensor modalities. Instead of implementing separate detection systems for each sensor type, it uses the same detection algorithm and object representation methods across LiDAR, radar, camera, and ultrasonic inputs, thereby managing complexity through algorithmic reuse rather than separate processing pipelines.
3Reliability
If safety actions are triggered based on multiple sensor confirmation, then the safety of occupants and pedestrians is improved, but the loss of time in response to actual hazards increases
Solution Approach 1:
The system performs preliminary detection and validation using multiple sensors before triggering safety actions. By continuously monitoring objects with redundant sensor inputs and pre-validating their presence through cross-sensor correlation, the system reduces false positives while maintaining rapid response capability. The preliminary action is the cross-verification process that occurs continuously in the background, ready to trigger immediate safety responses when confirmed hazards are detected.
Data Source
AI summary
Embodiments relate to the identification and mitigation of jamming attacks on autonomous vehicles. A technique includes determining that an object previously detected is no longer detected by a sensor of a vehicle. The technique includes, in response to the object previously detected no longer being detected by the sensor, determining that another sensor of the vehicle detects the object. The technique includes, in response to the object previously detected no longer being detected by the sensor, performing safety actions based on another sensor detecting the object.