1: \begin{abstract}

2: This paper presents a novel deep-learning-based approach to improve localizing radar measurements against lidar maps.

3: This radar-lidar localization leverages the benefits of both sensors; radar is resilient against adverse weather, while lidar produces high-quality maps in clear conditions.

4: However, owing in part to the unique artefacts present in radar measurements, radar-lidar localization has struggled to achieve comparable performance to lidar-lidar systems, preventing it from being viable for autonomous driving.

5: This work builds on ICP-based radar-lidar localization by including a learned preprocessing step that weights radar points based on high-level scan information.

6: To train the weight-generating network, we present a novel, stand-alone, open-source differentiable ICP library.

7: The learned weights facilitate ICP by filtering out harmful radar points related to artefacts, noise, and even vehicles on the road.

8: Combining an analytical approach with a learned weight reduces overall localization errors and improves convergence in radar-lidar ICP results run on real-world autonomous driving data.

9: Our code base is publicly available to facilitate reproducibility and extensions\footnote[2]{\,Paper code is available at: https://github.com/utiasASRL/mm\_masking.}.

10: \end{abstract}

11: