Abstract
Connected vehicles, whether equipped with advanced driver-assistance systems or fully autonomous, require human driver supervision and are currently constrained to visual information in their line-of-sight. A cooperative perception system among vehicles increases their situational awareness by extending their perception range. Existing solutions focus on improving perspective transformation and fast information collection. However, such solutions fail to filter out large amounts of less relevant data and thus impose significant network and computation load. Moreover, presenting all this less relevant data can overwhelm the driver and thus actually hinder them. To address such issues, we present Augmented Informative Cooperative Perception (AICP), the first fast-filtering system which optimizes the informativeness of shared data at vehicles to improve the fused presentation. To this end, an informativeness maximization problem is presented for vehicles to select a subset of data to display to their drivers. Specifically, we propose (i) a dedicated system design with custom data structure and lightweight routing protocol for convenient data encapsulation, fast interpretation and transmission, and (ii) a comprehensive problem formulation and efficient fitness-based sorting algorithm to select the most valuable data to display at the application layer. We implement a proof-of-concept prototype of AICP with a bandwidth-hungry, latency-constrained real-life augmented reality application. The prototype adds only 12.6 milliseconds of latency to a current informativeness-unaware system. Next, we test the networking performance of AICP at scale and show that AICP effectively filters out less relevant packets and decreases the channel busy time.
Introduction
Cooperative Perception (CP) enables vehicles to share sensor data (camera, LiDAR, etc.) via Vehicle-to-Everything (V2X) communication, effectively allowing them to "see" through the eyes of others and eliminate blind spots. However, this technology faces two critical bottlenecks: network congestion and driver information overload. Sharing raw or processed sensor data from numerous nearby vehicles can overwhelm wireless channels (DSRC or C-V2X). Furthermore, displaying every detected object on an Augmented Reality (AR) windshield can bury the driver under a sea of visual clutter, exceeding human cognitive capacity and potentially making driving more dangerous. AICP addresses these challenges by focusing on what is worth sharing rather than just how to share it.
System Design and Methodology
AICP employs a "full-stack" filtering approach that operates from the network layer up to the application layer. The architecture is designed for speed and relevance, ensuring that only the most informative data reaches the driver's display.
Key components include:
- Vehicular Data Unit (VDU): A lightweight data structure that encapsulates object metadata (position, velocity, type) for rapid interpretation.
- Contextual Multihop Routing (CMR): A network-layer protocol that filters packets based on relevance using directional routing and geographic limits, ensuring data is only forwarded to vehicles moving in similar directions within a useful range.
- Weighted Fitness Sorting: At the application layer, AICP ranks objects using a "fitness" score calculated via Mahalanobis distance. This metric accounts for the correlation between Distance, Velocity, Direction, and Category, allowing the system to prioritize high-risk objects (e.g., a nearby pedestrian) over irrelevant ones (e.g., a distant parked car).
Evaluation and Results
The system was evaluated using a real-world prototype in Helsinki for latency testing and a large-scale simulation of London city center for networking performance.
Performance and Efficiency
- Low Latency: The AICP pipeline adds only 12.6ms of overhead, keeping total end-to-end latency around 70ms—well within the 100ms safety threshold.
- Network Optimization: In high-density simulations, AICP reduced received Basic Safety Messages (BSMs) by 72% to 80% compared to standard flooding protocols, significantly reducing Channel Busy Time.
Visual Clarity
Qualitative results demonstrate a dramatic reduction in AR clutter. By prioritizing informative objects, the system highlights critical hazards while removing redundant or irrelevant bounding boxes that would otherwise distract the driver.
Conclusion
AICP represents a significant step forward in cooperative perception by treating informativeness as a first-class citizen. By filtering data at both the network and application layers, it successfully mitigates network congestion and prevents driver information overload without sacrificing safety-critical response times. Future work will focus on smoothing object persistence to prevent flickering and optimizing the human-computer interface for better peripheral placement of AR alerts.
Summary updated