CoLD Fusion: A Real-time Capable Spline-based Fusion Algorithm for Collective Lane Detection

Comprehensive environment perception is essential for autonomous vehicles to operate safely. It is crucial to detect both dynamic road users and static objects like traffic signs or lanes as these are required for safe motion planning. However, in many circumstances a complete perception of other objects or lanes is not achievable due to limited sensor ranges, occlusions, and curves. In scenarios where an accurate localization is not possible or for roads where no HD maps are available, an autonomous vehicle must rely solely on its perceived road information. Thus, extending local sensing capabilities through collective perception using vehicle-to-vehicle communication is a promising strategy that has not yet been explored for lane detection. Therefore, we propose a real-time capable approach for collective perception of lanes using a spline-based estimation of undetected road sections. We evaluate our proposed fusion algorithm in various situations and road types. We were able to achieve real-time capability and extend the perception range by up to 200%.

💡 Research Summary

The paper “CoLD Fusion: A Real-time Capable Spline-based Fusion Algorithm for Collective Lane Detection” addresses a critical bottleneck in autonomous driving: the inherent limitations of single-vehicle sensor ranges and visibility caused by occlusions and road curvatures. While previous research in Collective Perception (CP) has primarily focused on detecting dynamic objects like vehicles and pedestrians, this study innovatively extends CP to the detection of static, continuous structures—specifically, road lanes. The proposed “CoLD Fusion” algorithm utilizes Vehicle-to-Everything (V2X) communication to bridge the perception gap, effectively extending the lane detection range by up to 200%.

The methodology is bifurcated into two distinct driving scenarios: Convoy (Platooning) and General Driving. In the Convoy scenario, where the sensing ranges of the ego vehicle and a cooperative vehicle overlap, the authors implement a weighted average fusion strategy. By assigning a 75% weight to the cooperative vehicle and 25% to the ego vehicle, the algorithm leverages the higher accuracy of the leading vehicle, which is physically closer to the detected lane. In cases where no overlap exists, the algorithm simply concatenates the lane information from the lead vehicle.

For the General Driving scenario, which involves gaps in lane detection due to safety distances between vehicles, the authors propose a sophisticated four-stage spline-based fusion process. First, a “Fusion Feasibility Verification” step filters potential cooperative candidates based on strict spatial and directional error tolerances (±0.40m and ±10°, respectively). Second, to prevent the “flattening” effect—where splines lose curvature in bends—the “Appsis Estimation” step is introduced. This involves calculating an intersection point from extended lane endpoints and inserting a new control point at the 40% mark of the segment to maintain geometric fidelity. Third, “Spline Fitting” utilizes cubic splines to ensure $C^1$ and $C^2$ continuity, which is vital for smooth motion planning. Finally, the “Final Lane Construction” sequentially assembles the ego lane, the estimated spline, and the cooperative lane into a single, continuous lane model ($l_{coll}$).

Experimental results demonstrate that the algorithm achieves real-time performance while significantly expanding the perception horizon by 150% to 200%. However, the study acknowledges several limitations: the algorithm’s heavy reliance on precise localization (GPS/IMU), the omission of V2X communication latency and network congestion, and the assumption of a fixed 4-meter lane width. Future research directions include modeling uncertainty, implementing multi-vehicle simultaneous fusion, and developing hybrid approaches that integrate HD Maps with collective perception.