Map generation improvements

[popkarsd]

Current map generation node: (grid_map_generator.py in all_seaing_navigation)

• Written in python/rclpy, handles large amount of obstacle points via python lists
• Rasterizes obstacles from polygons found in unlabelled obstacle map (obstacle_map/raw)
• Adjusts probability of obstacle being in map using hard coded magic values to imitate bayesian inference

Desired map generation node:

• Written in C++/rclcpp to boost topic publish speed & decrease cpu usage
• Directly create obstacles from filtered point cloud data (point_cloud/filtered)
  • Use a cutoff on z-value to only retain point data in the plane of the boat
  • could cause issues becuz more noise?
• Actually use bayesian inference to adjust probability of obstacle being in map (not a priority) -> take a look at chapters 6 and 9 of "Probabilistic Robotics" (by Sebastian Thraun) on how to do that
  • Ensure points outside of lidar's pov are not updated