Authors: Khanh Nguyen, Thomas Walker, Hichem Telli, Carlos de la Torre-Ortiz, Stefano Fiorini, Salah Eddine Bekhouche, Mohammed-En-Nadhir Zighem, Wassim Swaileh, Tinghuai Wang
Organizations: Computer Vision Center, Universitat Aut`onoma de Barcelona, Spain · Independent Researcher · University of the Basque Country, Spain
High-fidelity LiDAR simulation remains a bottleneck for scalable development and validation of autonomous systems. Existing generators can synthesize realistic scans, but they are typically constrained to a single sensor configuration, offer limited control over structure and geometry, and drift in long-horizon 4D generation. We present CrossLiDAR, a unified transformer-based framework capable of generating controllable and spatio-temporally consistent multi-sensor LiDAR sequences. Our approach leverages a rectified flow objective with a Multi-Modal Diffusion Transformer (MM-DiT) backbone, within a compact latent space of range images. To support multiple LiDAR sensors, an Inter-sensor Module enables interaction across sensors, while a ControlNet-style branch provides structural conditioning (e.g., BEV priors or 3D bounding boxes). For modeling the temporal consistency, CrossLiDAR generates short N-frame sequences and extends them autoregressively to improve long-horizon coherence. Our model significantly outperforms state-of-the-art baselines, achieving an average improvement of 29.56% in JSD and 15.36% in MMD evaluated on nuScenes and the Waymo Dataset. CrossLiDAR bridges the gap to modern autonomous-driving simulation by enabling calibration-consistent multi-sensor LiDAR rollouts over long horizons for controllable, scenario-based testing.
Figures & tables
Figure 1: CrossLiDAR jointly generates synchronized LiDAR sequences for multi-sensor setups. Top: given multiple sensors mounted around the ego vehicle, CrossLiDAR synthesizes geometrically aligned per-sensor views (front/rear/left/right) that can be fused into a consistent ego-frame point cloud, while supporting structural conditions such as 3D bounding boxes. Bottom: our pipeline operates in a VAE-compressed range-image latent space; a rectified-flow module integrates an Inter-sensor Module for sensor-aware generation and a ControlNet branch for structural conditioning using BEV maps with projected 3D bounding boxes, yielding controllable multi-sensor LiDAR sequences.
Figure 2: Overview of CrossLiDAR Framework . Our framework utilizes a latent rectified-flow [ 25 ] formulation to generate synchronized multi-sensor LiDAR sequences. (Left) A frozen VAE encoder first maps range images into a compact latent space zt . (Right) These latents are processed by an MM-DiT [ 9 ] backbone, which incorporates two different attention mechanisms: an Inter-sensor Module for cross-sensor alignment and a LiDAR Sequence Module for temporal consistency. The generation is guided by a dedicated ControlNet [ 70 ] branch for BEV and 3D layout control. Finally, a frozen VAE decoder transforms the denoised latents back into high-fidelity temporal point clouds.
Method
Venue
Range
Points
BEV
FRD ↓
MMD ↓
FPD ↓
MMD ↓
JSD ↓
MMD ↓
Voxel
OpenDWM [ 36 ]
CVPR’25
—
—
7.55
52.23
16.25
2.29
OpenDWM-DiT [ 46 ]
CVPR’25
—
—
8.41
105.40
10.85
1.80
Range
LiDM [ 41 ]
CVPR’24
2.29
148.83
4.67
65.76
19.34
1.00
Table 1: Evaluations of Scene-level Fidelity on Waymo dataset for multi-sensor LiDAR generation.
Method
Venue
Range
Points
BEV
FRD ↓
MMD ↓
FPD ↓
MMD ↓
JSD ↓
MMD ↓
Voxel
UniScene [ 19 ]
CVPR’25
—
—
—
—
31.55
13.61
OpenDWM [ 36 ]
CVPR’25
—
—
0.95
5.43
20.17
5.61
OpenDWM-DiT [ 46 ]
CVPR’25
—
—
0.49
0.50
19.90
5.73
Range
Table 2: Evaluations of Scene-level Fidelity on nuScenes dataset for LiDAR generation.
Figure 3: Qualitative Results . The top row illustrates temporal sequence generation on nuScenes, where CrossLiDAR maintains strict temporal consistency and 3D bounding box alignment. The bottom row demonstrates multi-sensor Waymo generation, where a single unified scene is reconstructed by synchronizing outputs from four distinct generated sensors. Compared to baselines, CrossLiDAR produces scan patterns with improved scene-level geometric fidelity, effectively handling the complexities of multi-sensor generation.
Method
Venue
TTCE ↓
CTC ↓
Rollout ↓
1
2
1
2
3
4
CD
JSD
MMD
Voxel
OpenDWM [ 36 ] (t=5)
CVPR’25
990.17
990.74
0.37
0.52
0.65
0.68
2.81
0.53
0.57
OpenDWM-DiT [ 46 ] (t=5)
CVPR’25
1025.78
1026.43
0.39
0.59
0.76
0.81
0.78
0.13
0.34
Range-Temporal
LiDM [ 41 ] (t=3)
CVPR’24
1444.74
1444.48
0.20
0.24
—
—
0.53
0.52
1.81
Table 3: 4D LiDAR generation on Waymo using video metrics including: Temporal Consistency and Rollout-level metrics.
Method
Venue
TTCE ↓
CTC ↓
Rollout ↓
3
4
1
2
3
4
CD
JSD
MMD
Voxel
UniScene [ 19 ]
CVPR’25
2.74
3.69
0.90
1.84
3.64
3.90
—
—
—
OpenDWM [ 36 ]
CVPR’25
2.68
3.65
1.02
2.02
3.37
5.05
1.20
0.17
9.99
OpenDWM-DiT [ 46 ]
CVPR’25
2.71
3.66
0.89
1.79
3.06
4.64
0.68
0.17
8.80
Range-Temporal
Table 4: 4D LiDAR generation on nuScenes using video metrics including: Temporal Consistency and Rollout-level metrics. The temporal horizon is fixed to t=5 for all models.
Method
Venue
TTCE ↓
CTC ↓
Rollout ↓
Motion
3
4
1
2
3
4
CD
JSD
MMD
BVC ↑
FVD ↓
OpenDWM [ 36 ] (t=12)
CVPR’25
471.34
471.40
0.42
0.55
0.66
0.73
0.11
0.38
0.17
—
—
OpenDWM-DiT [ 46 ] (t=12)
CVPR’25
424.65
424.86
2.07
2.63
2.81
3.17
0.07
0.37
0.14
—
—
CrossLiDAR
—
369.63
369.74
0.05
0.07
0.08
0.09
0.07
0.12
0.02
0.79
57.42
Table 5: Auto-regressive generation on Waymo , including Temporal Consistency , Rollout-level and Motion-level metrics.
Method
Venue
TTCE ↓
CTC ↓
Rollout ↓
Motion
3
4
1
2
3
4
CD
JSD
MMD
BVC ↑
FVD ↓
FVMD ↓
UniScene [ 19 ] (t=5)
CVPR’25
2.74
3.69
0.90
1.84
3.64
3.90
—
—
—
—
—
—
OpenDWM [ 36 ] (t=16)
CVPR’25
2.64
3.16
0.82
1.62
2.83
4.67
0.97
0.46
0.27
0.22
3923.11
4147.61
OpenDWM-DiT [ 46 ] (t=16)
CVPR’25
1.23
1.39
0.68
1.42
2.42
3.83
0.83
0.24
0.13
0.43
2835.68
2261.98
LiDARCrafter [ 22 ] (t=5)
AAAI’26
2.65
3.56
1.12
2.38
3.02
4.81
—
—
—
—
—
—
CrossLiDAR (t=16)
—
0.41
0.50
1.09
1.22
1.36
1.47
0.47
0.22
0.26
0.88
307.07
1672.90
Table 6: Autoregressive generation on nuScenes , including Temporal Consistency , Rollout-level and Motion-level metrics.
Sensor
VFOV
HFOV
Resolution
Range (m)
Intensity
Velodyne HDL-32E
[+10∘;−30∘]
360∘
(32, 1080)
[2.0;100.0]
[0;225]
Table 7: LiDAR Sensor Specifications for the nuScenes Dataset . We report the vertical field of view ( VFOV ), horizontal coverage ( HFOV ), range-image resolution (rows × columns), valid range (m), and intensity range.
Sensor
VFOV
HFOV
Resolution
Range (m)
Intensity
TOP
[−17.6∘,+2.4∘]
360∘
64×2650
[1.2,75.0]
[7×10−6,4×106]
F/SL/SR/R
[−90∘,+30∘]
—
200×600
[5×10−3,20.0]
[5×10−6,40.5]
Table 8: LiDAR Sensor Specifications on Waymo Open Dataset . We report vertical field of view (VFOV), horizontal coverage (HFOV), range-image resolution, valid range (m), and intensity range for TOP and SIDE sensors.
Figure 4: Waymo range-image examples. Top: TOP LiDAR range image ( 64×2650 ), with 360∘ horizontal coverage. Bottom: four SIDE LiDAR range images ( 200×600 each), with higher vertical resolution and partial-view coverage.
Figure 5: Examples of Waymo point-cloud coverage. Left : Point cloud from each SIDE LiDAR sensor, illustrating the partial horizontal coverage provided by each individual sensor. Right : Point cloud obtained by fusing the four SIDE LiDAR sensors, shown in comparison with the TOP LiDAR. The SIDE sensors are color-coded by position: front ( yellow ), left ( pink ), rear ( blue ), and right ( green ).
TrainVal
Test
Vehicle
941.6K
227K
Pedestrian
499.1K
121.4K
Cyclist
11.8K
3.4K
Total
1451.6K
352K
Table 9: Statistics on annotations.
Train
Val
Test
Scenes
798
202
150
Frames
158K
40K
30K
Clips
21K
5K
5K
Table 10: Size of the data splits.
Figure 6: Per-frame annotations.
Figure 7: Densification on the Waymo SIDE LiDAR range images . The top row shows the original sparse range images from Waymo, while the bottom row shows the densified results. The densification algorithm [ 18 ] fills in missing pixels and significantly enhances the visual quality and completeness of the range images.
Figure 8: Rollout examples of temporal consistency on nuScenes and Waymo. For each dataset, we visualize a scene composed of multiple point clouds acquired at different timestamps. The consistent alignment of the colored regions across time indicates that the reconstruction preserves temporal consistency. The highlighted objects include a vehicle ( orange ), a pedestrian ( blue ), vegetation ( green ), and background structures ( purple ).
Method
Compression
Input Size
Transform
Variant
R-FSVD ↓
R-FPVD ↓
CD (10−3)↓
EMD (10−3)↓
LiDM [ 41 ]
fc=4,fp=4†
208 × 640
LogNorm
—
426.71
191.74
505.98
102.68
fc=4,fp=4†
208 × 640
LogNorm
w/o LDesc
1.83
0.32
2.81
16.25
RangeLDM [ 13 ]
fp=8†
224 × 608
LogNorm
—
3.94
0.31
3.43
23.96
fp=8†
224 × 608
LogNorm
w/o LDesc
1.26
0.19
0.98
7.39
fp=8†
200 × 600
LogNorm
—
3.55
0.27
1.27
13.11
fp=8†
200 × 600
LogNorm
w/o LDesc
1.44
0.15
0.49
4.40
Table 11: VAE ablation results on Waymo SIDE LiDAR. Reconstruction quality is compared across VAE backbones and normalization strategies. Configuration † : compression rate f=64 , range densification, and filling invalid points. LDesc denotes the discriminator (adversarial) loss. Performance is evaluated using perceptual metrics: R-FSVD, R-FPVD and statistical metrics: Chamfer Distance (CD), Earth Mover’s Distance (EMD), following [ 41 ] . The best result is highlighted in bold and the second-best in underline .
Figure 9: Impact of Logarithmic Transformation on LiDAR range distributions. Left: For TOP LiDAR, the log transform significantly reduces the skewness of high dynamic range values. Right: For SIDE LiDAR on Waymo, where the range values are less skewed and cover a shorter distance, the effect of the log transform is minimal.
Figure 10: Effects of range image resizing on generated point clouds. Resizing the range image from (200,600) to (224,608) introduces noise and geometric distortions in the resulting point cloud, noticeably affecting object shapes (e.g., trees, cars). Best viewed when zoomed in .
Method
Input Size
Backbone
Range
Points
BEV
FRD ↓
MMD ↓
FPVD ↓
MMD ↓
JSD ↓
MMD ↓
LiDM † [ 41 ]
208 × 640
U-Net
2.29
148.83
4.67
65.76
0.19
1.00
200 × 600
DiT
0.45
77.91
2.52
113.33
0.15
1.00
RangeLDM † [ 13 ]
224 × 608
U-Net
10.97
979.33
122.38
1236.59
0.22
1.63
200 × 600
DiT
9.07
938.83
61.47
374.81
0.19
1.16
CrossLiDAR † (t=5)
200 × 600
DiT + Naive-Stacking
1.66
55.67
14.82
156.55
0.20
1.62
Table 12: Ablations on Diffusion Backbone on Waymo SIDE LiDAR . Configuration † : compression rate f=64 , range densification, filling invalid points and without discriminator. The best result is highlighted in bold and the second-best in underline .
Venue
Modality
LiDAR Sensors
LiDAR Frames
Supported Conditions
Diffusion Model
Dataset
Code
Checkpoint
LiDAR Diffusion [ 41 ]
CVPR’24
LiDAR
1
1
image,semantic map
U-Net
K
Yes
K
RangeLDM [ 13 ]
ECCV’24
LiDAR
1
1
inpaint,downsample
U-Net
K,N
Yes
K,N
R2DM [ 35 ]
ICRA’24
LiDAR
1
1
—
U-Net
K
Yes
K
LiDARDM [ 74 ]
ICRA’25
LiDAR
1
multi
map
U-Net
K,W
Yes
K,W
UniScene [ 19 ]
CVPR’25
LiDAR &Camera
1
multi
BEV → semantic occupancy
Sparse U-Net
N
Yes
N
OpenDWM [ 36 ]
CVPR’25
LiDAR
1
multi
BEV,map,text
U-Net/DiT
N
Yes
N
Table 13: Summary of existing methods for LiDAR generation. Overview of compared methods, indicating sensor setup, generative model, available modalities, supported input conditions, training datasets ( K : KITTI-360 [ 23 ] , N : nuScenes [ 4 ] , W : Waymo [ 49 ] ), and code/checkpoint availability.
Figure 11: Naive sensor stacking for multi-sensor LiDAR generation. Each sensor range image at time step t is independently encoded by the VAE. The diffusion model operates on the stacked multi-sensor latents, enabling generation (or prediction) of joint sensor, and can be extended to sequences by stacking along the temporal dimension.
Figure 12: Cross-Sensor Alignment on Waymo. Qualitative comparison between our method (the last two rows) and competing approaches (the first row). While competitors generate each sensor stream independently, leading to misalignment and geometric inconsistencies across viewpoints, our Inter-Sensor Module enables coherent multi-view generation with consistent object geometry and spatial alignment across all LiDAR sensors.
Figure 13: Temporal Consistency and Controllable Generation on Waymo. Qualitative results demonstrating both temporal coherence and controllable generation. The first column shows BEV and 3D-box conditioning signals, while the remaining columns show generated LiDAR frames at different timesteps. Across consecutive frames, CrossLiDAR preserves stable object geometry and motion consistency while adhering to structural control signals for object placement and scene layout.
University of Science and Technology of China, Hefei, Anhui 230026, China · Shanghai Innovation Institute, Shanghai 200231, China · Fudan University, Shanghai 200433, China +1
College of Computer Science and Artificial Intelligence, Fudan University, Shanghai, China · School of Computing, Department of Computer Science, National University of Singapore, Singapore · College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, China +5