How to use NDT and ICP matching on the Moriyama dataset in Python with Open3D?

Question

I am new to the area of autonomous vehicles and I am having trouble implementing NDT and ICP matching in Python on the Moriyama dataset. Specifically, I have the sample map data (downloaded .pcd files) and sequence data from the ROS bag (4700+ .pcd files). However, when trying to register the sequence points to the map, I find the the map is not visible and I get a lot of [Open3D WARNING] [ReadPCDData] Failed to read data record. [Open3D WARNING] Read PCD failed: unable to read data. [Open3D WARNING] [KDTreeFlann::SetRawData] Failed due to no data. However, I can visualize each piece individually.

I simplified the problem by just trying to visualize the map first before registering any points. I was able to successfully piece together the map .pcd files and I can visualize the map in Python with Open3D. The related code and image of the map is below:

import open3d as o3d

# Path to the map point cloud file
map_file_path = "./data/combined_source.pcd"

# Read the point cloud file
map_point_cloud = o3d.io.read_point_cloud(map_file_path)

# Create a visualizer
visualizer = o3d.visualization.Visualizer()
visualizer.create_window(window_name='Open3D Map Visualization', width=800, height=600)

# Add the map point cloud to the visualizer
visualizer.add_geometry(map_point_cloud)

# Set the view control (you can adjust this based on your preference)
view_control = visualizer.get_view_control()
view_control.set_lookat([0, 0, 0])
view_control.set_up([0, 1, 0])
view_control.set_front([1, 0, 0])
view_control.set_zoom(0.8)

# Run the visualizer
visualizer.run()

# Destroy the visualizer window
visualizer.destroy_window()

Map of Moriyama Generated from above Code

The image may look a bit funny, but zooming in on the pop up window when it is generated reveals that this is indeed the map of Moriyama. The issue still lies in registering the sequence data to the map though. I would like to apply NDT and ICP matching with the sequence data that I have (4700+ .pcd files). An approach I've thought of is iteratively registering one point at a time to the map, but that didn't seem to work either (or may not have implemented properly). Could I get assistance with this issue?