SELFY project aims at pushing towards a safer, more resilient and cybersecure CCAM ecosystem. To do so, vehicles and infrastructure have to cooperate, especially to build situational awareness and collective perception. In this blog post, we will tell you about SELFY infrastructure model.
Complementary sensors for accurate perception
To perceive local environment, SELFY infrastructure relies on different types of sensors, such as cameras, lidars and radars. The data obtained from each sensor is then processed to extract a list of objects present in the scene. Video frames are analysed by the Traffic Monitoring Tool (TMT), whereas point clouds are analysed by the Sensor Fusion & Anomaly Detection Tool (SFAD).
Once objects have been extracted for each sensor, the SFAD gathers them and performs a fusion operation. The fact that different sensors are used enables improving the accuracy of the situational awareness. Indeed, using multiple sensors, the scene can be seen from different points of view, which allows avoiding issues like occlusions.
In addition, using different types of sensors is also useful as each type of sensor has its own advantages: while cameras, complemented with deep learning, are very good at determining what kind of objects are present, lidars allow very precise positioning of objects. As a result, the SFAD benefits from the advantages of each sensor when fusing results.
A critical eye for trustworthiness
In an ideal world, the SFAD would deliver a perfectly accurate representation of the environment. However, the world is not ideal, and SELFY has a strong focus on security and trustworthiness. What if a camera or a lidar is tampered or hacked?
Such issues are addressed through the analysis of the SFAD results by another key component of SELFY infrastructure: the Threat Evaluation Tool (TET). The Threat Evaluation Tool gathers data obtained from SFAD and searches for suspicious patterns. For instance, if all sensors but one repeatedly agree on the content of a scene, there is probably an issue. The Threat Evaluation Tool detects such issues and reports them, so that investigations can be made.
What about V2X?
SELFY project is not about infrastructure alone, but more generally about CCAM. Hence, SELFY infrastructure has a Road Side Unit (RSU) to exchange messages with other V2X agents, especially V2X-enabled vehicles.
When a nearby car emits a Cooperative Awareness Message (CAM) to describe its current position, infrastructure receives this message and forwards it to the SFAD to check whether it is consistent with infrastructure own perception. If not, either the vehicle has a problem (e.g. its localization means do not work), or the V2X agent is purposefully misbehaving. In both cases, if the inconsistency persists, the V2X agent will be reported for further investigations.
SELFY infrastructure identifies emitters of inconsistent V2X messages, either due do malfunctions or to attacks.
SELFY infrastructure does not only receive V2X messages: based on its own perception of the environment, it also sends Collective Perception Messages (CPM) that provide a description of its situational awareness to nearby V2X agents.
As a result, any vehicle that receives said messages can take advantage of them to increase its own awareness. For instance, this may allow a vehicle to warn the driver about a coming bicycle that cannot yet be perceived by the vehicle own sensors.
Pushing towards a safer, more secure CCAM
By combining different sources of data and analysing them with a critical eye, SELFY infrastructure is designed to get an accurate and trustworthy understanding of its environment.
This situational awareness benefits all V2X agents, by making them aware of things they cannot perceive on their own. Finally, this design also allows identifying malfunctions and misbehaviours, which contributes to raising the safety and security of CCAM.
Author: Canon Research Centre France
