Generating shard data from a geographical area split in sections

One or more non-transitory computer-readable storage media comprising instruction for execution which, when executed by a processor, are operable to perform operations for generating shard data, the operations comprising:

• loading data representative of a geographical area that has been split into guiding sections, wherein the guiding sections are generated based on an algorithm that optimizes the guiding sections based on memory usage by clients, data storage per resulting shard file, and object density within a section;
• loading data representative of semantic objects in the geographical area;
• searching, for each of the guiding sections, for semantic objects that are located at least partly in a section;
• applying rules defining which semantic objects are to be grouped together into a single shard data entry and which semantic objects are to be stored in separate shard data entries, wherein the rules comprise:
• (1) a lane and boundaries of the lane are to be grouped together;
• and (2) lanes and intersection of the lanes are to be grouped together;
• generating a shard data entry per guiding section, the shard data entry identifying one or more found semantic objects that were found by the searching and fulfilled the rules;
• computing a bounding box per shard data entry based on the one or more found semantic objects of the shard data entry, wherein the bounding box is indicative of the geographical area covered by the shard data entry and is different from the guiding section that was used in the searching;
• generating and storing shard files and a metadata file using the shard data entries and the bounding boxes;
• and
• causing at least one of the shard files to be delivered to at least one of the clients using the metadata file, wherein the at least one of the clients include an autonomous vehicle that makes a driving decision based on the at least one of the shard files.

A computer system for generating and delivering shard files to clients, the computer system comprising:

• a shard-to-object relations database;
• a shard metadata database;
• one or more file storage;
• one or more processors to execute instructions that cause the one or more processors to:
• split a geographical area into guiding sections based on memory usage by the clients, data storage per resulting shard file, and object density within a section;
• identify, for each of the guiding sections, for semantic objects that are located at least partly in a section;
• apply rules defining which semantic objects are to be grouped together into a single shard data entry and which semantic objects are to be stored in separate shard data entries, wherein the rules comprise:
• (1) a lane and boundaries of the lane are to be grouped together;
• and (2) lanes and intersection of the lanes are to be grouped together;
• generate a shard data entry per guiding section, the shard data entry identifying one or more found semantic objects that were found by the identifying and fulfilled the rules;
• determine a bounding box per shard data entry based on the one or more found semantic objects of the shard data entry, wherein the bounding box is indicative of the geographical area covered by the shard data entry and is different from the guiding section that was used in the identifying;
• store the shard data entries in the shard-to-object relations database;
• store the bounding boxes in the shard metadata database;
• generate and store shard files and a metadata file using the shard data entries and the bounding boxes in the one or more file storage;
• and
• cause at least one of the shard files to be delivered to at least one of the clients using the metadata file, wherein the at least one of the clients include an autonomous vehicle that makes a driving decision based on the at least one of the shard files.