Cooperative, connected, and automated mobility (CCAM) is revolutionizing the way we think about transportation. These advanced systems promise to enhance road safety, reduce traffic congestion, and minimize environmental impact. Imagine a world where vehicles communicate seamlessly with each other and their surroundings, making split-second decisions to optimize travel efficiency and safety. The potential benefits are immense, from fewer accidents to smoother commutes and lower emissions.
But what happens if something fails? In a world increasingly reliant on technology, the possibility of system failures cannot be ignored. Whether it’s a sensor malfunction, a communication breakdown, or a cyberattack, the stakes are high. This is where fallback strategies come into play. These strategies are designed to ensure that, even in the face of unexpected failures, Connected and Automated Vehicles (CAVs) can maintain safety. Within the SELFY project, the Cooperative Resilience and Healing System (CRHS) team is working on methods to overcome localization failures.
Localization is crucial for CAVs as it allows them to accurately determine their position on the road. This capability is essential for safe navigation, precise maneuvering, and effective decision-making. Without reliable localization, CAVs cannot maintain lane discipline, avoid obstacles, or follow traffic rules, which compromises safety and efficiency. To localize accurately, CAVs rely on a combination of sensors, including LiDAR, GNSS, and cameras. Each of these sensors plays a crucial role, but they also come with their own set of vulnerabilities.
- LiDAR (Light Detection and Ranging) uses laser pulses to detect the vehicle’s surroundings. This technology is highly accurate and effective in various weather conditions. However, LiDAR’s effectiveness is contingent on having a prebuilt map of the environment.
- GNSS (Global Navigation Satellite System) provides positioning data by communicating with satellites. While GNSS is essential for determining a vehicle’s location, it is not without its flaws. GNSS signals can be disrupted by physical obstructions like tall buildings or tunnels, leading to signal loss. Additionally, GNSS is vulnerable to cyberattacks, such as spoofing or jamming, which can compromise the accuracy and reliability of the positioning data.
- Cameras are used to capture visual information about the vehicle’s environment. Apart from detecting and recognizing objects, road signs, and lane markings, cameras enable the use of algorithms known as Visual Odometry (VO) to localize the CAV. However, cameras are highly dependent on lighting conditions. Poor visibility due to darkness, fog, or glare can significantly impair their performance, making it challenging for the vehicle to accurately localize itself.
What are we doing within the SELFY project?
Imagine a situation in which the CAV is driving in a map-less zone with poor visibility conditions for cameras, relying only on the GNSS for localization. Now imagine that all of a sudden, the GNSS fails or reduces its accuracy. As you can imagine, it is complicated to continue the normal operation of the vehicle, but safety cannot be jeopardized.
To overcome this possible situation, we are developing a fallback localization algorithm that uses the detection of road objects (landmarks), such as other vehicles, to localize the vehicle using relative measurements. If we know the position of at least one object before the GNSS failure, and we continue detecting it, the relative detections can help us to localize our new position. What is more, we can continue the operation as long as we continue detecting new objects before losing the visibility of the previous ones.
As you can imagine, this algorithm is designed to work as a fallback mechanism and is not suitable for being used for more than a few seconds, because it stops working the moment every landmark is lost. Consequently, it should be used only to execute a safe stop maneuver.
In the video, you can see it in action: Once the GNSS is deactivated (simulated with the bottom left command), the algorithm starts working. There are three different coordinate axes you should notice. The first one is static and marks the last well-known position of the CAV (when the GNSS was available). Once the vehicle starts moving, two more coordinate axes appear. The more stable one is the ground truth, this is, the exact position of the vehicle (known because it is a simulation). Finally, the noisy coordinate axis corresponds to the position calculated by the algorithm. As you can see, it is not perfect, but combined with a filter could secure a safe stop manoeuvre when needed.
The algorithm will be explained in detail and performance metrics will be given in different scenarios in future scientific publications from the SELFY team.
Author: Tecnalia
