US 12,338,072 B1
System and/or method for dynamic rebalancing of multi-rack storage
Christopher Walti, South San Francisco, CA (US); Ahmad Baitalmal, South San Francisco, CA (US); Jefferson Packer, South San Francisco, CA (US); Michael Brevoort, South San Francisco, CA (US); and Allen Ferrick, South San Francisco, CA (US)
Assigned to Mytra, Inc., South San Francisco, CA (US)
Filed by Mytra, Inc., South San Francisco, CA (US)
Filed on Nov. 29, 2024, as Appl. No. 18/964,303.
Claims priority of provisional application 63/604,069, filed on Nov. 29, 2023.
Int. Cl. B65G 1/04 (2006.01); B25J 9/16 (2006.01); B65G 1/06 (2006.01); B65G 1/137 (2006.01); G05B 13/02 (2006.01)
CPC B65G 1/137 (2013.01) [B25J 9/1664 (2013.01); B65G 1/0414 (2013.01); B65G 1/0471 (2013.01); B65G 1/0478 (2013.01); B65G 1/06 (2013.01); B65G 1/065 (2013.01); G05B 13/027 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system for automated storage and automated retrieval (AS/AR) comprising:
a frame structure defining a three-dimensional (3D) rectilinear grid of cells;
a plurality of robots within the frame structure configured to translate payloads between cells of the 3D rectilinear grid;
a motion planner communicatively coupled to each of the plurality of robots and configured to provide instructions to each robot of the plurality of robots based on a target state of each robot of the plurality of robots and payload within the frame; and
a management system comprising a set of models, the set of models configured to determine the target state with a set of inputs comprising:
a position and a set of attributes for each indexed payload of a plurality of indexed payloads;
a set of predefined constraints of the rectilinear grid;
a set of robot motion constraints; and
a set of historic operation data.
 
11. A method for a set of automated storage and automated retrieval (AS/AR) robots comprising:
receiving a set of inputs from the set of AS/AR robots within a multi-rack frame structure, the multi-rack frame structure defining a three-dimensional (3D) rectilinear grid of translation axes;
based on the set of inputs, determining a current state of a warehouse, the state comprising a position of a plurality of payloads relative to the 3D rectilinear grid;
based on the current state, determining a target state using a set of models comprising a neural network model pretrained by reinforcement learning (RL);
at a centralized motion planner, determining instructions for each AS/AR robot of the set of AS/AR robots based on the target state; and
dynamically controlling the set of AS/AR robots based on the instructions.