CLONeR: <b>C</b>amera-<b>L</b>idar Fusion for <b>O</b>ccupancy Grid-Aided <b>Ne</b>ural <b>R</b>epresentations

Recent advances in neural radiance fields (NeRFs) achieve state-of-the-art novel view synthesis and facilitate dense estimation of scene properties. However, NeRFs often fail for outdoor, unbounded scenes that are captured under very sparse views with the content concentrated far away from camera, as is typical field robotics applications. In particular, NeRF-style algorithms perform poorly: 1) when there insufficient little pose diversity, 2) contain saturation shadows, 3) finely sampling large fine structures becomes computationally intensive. This letter proposes CLONeR , which significantly improves upon NeRF by allowing it to model outdoor driving observed input sensor views. achieved decoupling occupancy color learning within framework into separate Multi-Layer Perceptrons (MLPs) trained using LiDAR camera data, respectively. addition, this a method build differentiable 3D Occupancy Grid Maps (OGM) alongside model, leverage grid improved points along ray volumetric rendering metric space. Through extensive quantitative qualitative experiments on KITTI dataset, demonstrates proposed outperforms models both depth prediction tasks data.