System for training and deploying generative language model for formulating instructions for facility asset and updating different map types of facility

A system, comprising:

• two or more stationary assets, each of the two or more stationary asset comprises a tag, a transmitter, and a receiver for communicating a signal with the tag;
• a dynamic asset comprising a vehicle;
• at least one memory;
• and
• at least one processor configured to:
• receive natural language input corresponding to a task to be performed by the dynamic asset within a facility;
• determine, using a Machine Learning (ML) model, first instructions for the dynamic asset to perform the task, wherein determining the first instructions comprises applying the natural language input as a first input to the ML model, wherein the first instructions comprise a first location to which the dynamic asset is to move;
• transmit the first instructions to the dynamic asset via a network;
• execute the first instructions by the dynamic asset to perform the task comprising moving a current location of the dynamic asset;
• receive, by the ML model, feedback information comprising a metric determined using the current location of the dynamic asset, wherein the current location is determined based on a distance between the tag and each of the two or more stationary assets, using a location of each of the two or more stationary assets and a Time-of-Flight (ToF) of the signal communicated between the tag and each of the two or more stationary assets;
• update the ML model using the current location of the dynamic asset using reinforced learning, comprising:
• determining a dynamic asset metric based on a distance between the current location of the dynamic asset and a location in the first instructions;
• determining an efficiency metric using a total distance traveled by the dynamic asset within a predetermined time period by aggregating incremental distances between two locations of the dynamic asset sampled according to a time granularity over the predetermined period;
• determining an assignment verification metric comprising verifying whether a semi-dynamic asset being moved by the dynamic asset is appropriate by verifying that a distance between the current location of the dynamic asset and a current location of the semi-dynamic asset being within a predetermined range;
• determining a time-to-completion metric according to a difference between a first timestamp indicating the start of executing the first instructions and a timestamp indicating the end of performing the first instructions;
• determining an idle-time metric according to a time duration within which the dynamic asset is not executing any instructions;
• defining a reward framework comprising indication of positive performance and negative performance based on the metric comprising an aggregate of the dynamic asset metric, the efficiency metric, the assignment verification metric, the time-to-completion metric, and the idle-time metric;
• and
• updating the ML model using the reward framework;
• determine, using the updated ML model, second instructions for the dynamic asset to perform the task, wherein the second instructions comprise a second location to which the dynamic asset is to move;
• transmit the second instructions to the dynamic asset;
• execute the second instructions by the dynamic asset to perform the task comprising moving a current location of the dynamic asset;
• generate a first map of the facility including the current location of the dynamic asset, wherein the first map comprises, in a 2-Dimensional (2D) or 3-Dimensional (3D) coordinate system, a current location of each of a plurality of stationary assets of the facility, a current location of each of a plurality of semi-dynamic assets of the facility, and a current location of each of a plurality of dynamic assets of the facility, wherein the plurality of dynamic assets comprises the dynamic asset;
• update a second map and a third map based on the first map, the second map being generated by the dynamic asset using at least one sensor thereon, the dynamic asset comprises a vehicle, the second map comprises a Simultaneous Localization and Mapping (SLAM) map, and the third map comprises a Computer-Aided Design (CAD) drawing, wherein updating the second map and the third map comprises:
• translate coordinates of the dynamic asset defining the second location in the first map to coordinates of the dynamic asset defining the second location in the second map;
• and
• translate coordinates of the dynamic asset defining the second location in the first map to coordinates of the dynamic asset defining the second location in the third map.