Walmart Apollo, LLC

Mingang Fu

Jing Huang

Madhavan Kandhadai Vasantham

Minghui Liu

Amritayan Nayak

Deepak R Deshpande

Shengyang Zhang

Aditya Arcot Srinivasan

Prakash Seetharaman

Devadas Pattathil

Seon Ki Kim

Ou Sun

Joseph Wayne Hendricks

Chenxin Ma

Vidyanand Krishnan

Rohit Jain

Arun Prasad Nagarathinam

Pushkar Raj Pande

Surnaik Prakash Srivastava

Kunlei Lian

Jamie Lee Hamilton

Jennifer Chen

Apeksha Mehta

Pawan Kumar

Li Ji

Ankit Jasuja

Ti Zhang

Anurag Gupta

Srinivasan Venkatesan

Syed Aman

Kumar Malyala

Liqing Zhang

Yu Zheng

Ming Ni

Xiaojie Wang

Tiantian Nie

Deyi Zhang

Ranjith Moola

Bipin Mulik

Akshara Dendi

A method including obtaining an optimization request at a coordinating engine. The method also can include triggering engines to process the optimization request. At least one of the engines divides the optimization request into subproblems. At least a portion of the engines solve the subproblems. Respective instances of at least one of the engines are triggered to handle respective ones of the subproblems. Each of the engines provides a dynamic algorithmic flow using modularized algorithmic solvers. The dynamic algorithmic flow is adjusted based on a respective input to each of the engines. The method additionally can include outputting, from the coordinating engine, one or more results in response to the optimization request, based on results for the subproblems generated by the engines. Other embodiments are described.

Divide-and-conquer framework and modularized algorithmic scheme for large-scale optimization

A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform operations: receiving a load design for loading a trailer to deliver orders; generating an initial grouping comprising stacks of pallets that are configured to be loaded interchangeably among respective floor spot assignments on the trailer; iteratively adjusting at least one of the stacks of pallets within the initial grouping to satisfy at least a center-of-gravity constraint in a final load design for the trailer; and outputting the final load design comprising the initial grouping to cause the stacks of pallets of the initial grouping to be loaded into the trailer according to the final load design. Other embodiments are described.

Automatic generation of flexible load design

A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions, that when executed on the one or more processors, cause the one or more processors, to perform: obtaining a first load plan, wherein the first load plan comprises a set of stacks assigned to floor spots of a trailer; determining that at least one of (a) there is an empty floor spot in the first load plan for the trailer or (b) the first load plan is infeasible; determining a target number of stacks for an updated load plan; iterating, via simulated annealing, through modifications to the updated load plan; terminating the iterating when a predetermined termination criteria is satisfied; and outputting the updated load plan. Other embodiments are disclosed.

Unstacking optimization

This application relates to apparatus and methods for generating and implementing a machine learning model in electronic delivery systems to determine delivery routes and times. In some examples, a computing device generates a machine learning model comprising a plurality of indexed binary trees. Each indexed binary tree determines either a first value, or a second value, based on comparing an input to a condition value. The machine learning model can generate prediction values based on the determined values of all indexed binary trees. The machine learning model is trained with historical data. Once trained, the machine learning model's performance is evaluated. Based on the evaluation, the machine learning model may be further refined. Once the machine learning model's performance satisfies requirements, the computing device employs the machine learning model to determine vehicle delivery routes and estimated delivery times.

Apparatus and methods for determining delivery routes and times based on generated machine learning models

A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform: determining that a respective number of totes for a respective order of one or more orders can be reduced; reducing, using an infeasible totes loop, the respective number of the totes for the respective order to a respective reduced number of the totes; executing, as part of the infeasible totes loop, an item search; running the item search until a predefined number of iterations is reached; generating directions for a route of a pick walk based on a map of a retail facility; and facilitating displaying, on an electronic device, the route of the pick walk on the map of the retail facility. Other embodiments are disclosed herein.

Systems and methods for optimization of pick walks

A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform certain operations. The operations can include determining an estimated arrival time of a user at a physical store. The operations also can include generating an estimated wait time using a machine learning model and based on input data comprising the estimated arrival time and dynamic wait time data for the physical store. The operations additionally can include sending the estimated wait time to at least one of the physical store or a mobile device of the user. Other embodiments are disclosed.

Automatic determination of pickup wait times

A method including obtaining information about a trailer that has been partially loaded with preloaded stacks in a manner that deviates from an original load design. The trailer is loaded with stacks of pallets including (i) the preloaded stacks that have already been loaded in the trailer and (ii) unloaded stacks that have not yet been loaded into the trailer. The method also can include determining positions of empty floor spots remaining in the trailer. The method additionally can include determining a first portion of an incremental load design for the unloaded stacks using a gap-filling pattern behind an uneven rear edge of the preloaded stacks in the trailer. The method further can include determining a second portion of the incremental load design. The method additionally can include updating the incremental load design based on an overall load design of the trailer using a first simulated annealing using a first neighborhood defined by separate rows within a delivery group that does not include the preloaded stacks. The method further can include outputting at least the incremental load design, as updated, to cause the unloaded stacks to be loaded in the trailer according to the incremental load design while the preloaded stacks remain in the trailer. The incremental load design specifies a respective floor spot assignment for each of the unloaded stacks. Other embodiments are described.

Automatic generation of incremental load design

A method including setting a preferred delivery route as an initial delivery route comprising original delivery stops. The method further can include evaluating, in real-time, the preferred delivery route, comprising: (a) determining a to-be-resequenced delivery stop list by moving a random quantity of delivery stops from the original delivery stops to the to-be-resequenced delivery stop list based on a delivery-stop-selecting act, wherein: after the to-be-resequenced delivery stop list is determined, a modified preferred delivery route comprises remaining delivery stops from the original delivery stops; (b) after determining the to-be-resequenced delivery stop list, creating, in real-time, a resequenced delivery route by adding each delivery stop of the random quantity of delivery stops of the to-be-resequenced delivery stop list to the modified preferred delivery route; (c) determining, in real-time, whether the resequenced delivery route is preferred over the preferred delivery route; and (d) when the resequenced delivery route is determined to be preferred over the preferred delivery route, resetting, in real-time, the preferred delivery route to be the resequenced delivery route. The method further can include transmitting, in real-time via a computer network, the preferred delivery route to a driver mobile device of a delivery driver for the delivery driver to follow the preferred delivery route to make deliveries. Other embodiments are disclosed.

Automatically planning delivery route

A system includes one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform: selecting, by a greedy tote reduction algorithm within an infeasible totes loop, items that have a highest volume restriction and a highest weight restriction for each of a plurality of respective totes; iteratively executing a swap of the items; iteratively searching for candidate solutions until a solution of picklists for the plurality of respective totes containing the items, as swapped, is found; executing a minimum trolley loop algorithm on the solution of picklists to create combined picklists; and displaying to a picker on an interface of a computing device, turn-by-turn directions within a pick walk for the combined picklists. Other embodiments are disclosed herein.

A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions configured to run on the one or more processors and perform: generating a feasible dock-out plan designed to deliver one or more orders to one or more physical stores based on a feasible route plan by: generating a preliminary dock-out time boundary based on one or more delivery time windows; determining an earliest dock-out time based at least in part on the preliminary dock-out time boundary and a backward search; and determining a latest dock-out time based at least in part on the preliminary dock-out time boundary and a forward search; and transmitting a range of dock-out times comprising (i) the earliest dock-out time and (ii) the latest dock-out time to a user. Other embodiments are described.

Flexible dock-out time

Systems and methods including one or more processors and one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and cause the one or more processors to perform receiving one or more orders comprising a list of one or more pallets; grouping each pallet of the one or more pallets into one or more groups using one or more respective qualities of each respective pallet; for each group of the one or more groups, building one or more stacks of pallets using at least three different stacking models; selecting a final stack list of a model of the at least three different stacking models having a lowest number of stacks; and optimizing the final stack list using a probabilistic model. Other embodiments are disclosed herein.

Systems and methods for vehicle routing

A method including planning delivery routes using clustering. The method can include generating clusters for nodes for order data based, at least in part, on (a) location information of the nodes and (b) load capacity information of delivery vehicles. The method also can include determining a respective centroid for each of the clusters. Further, the method can include computing a node-cluster distance matrix. Additionally, the method can include reassigning the nodes to the clusters based, at least in part, on: (a) the node-cluster distance matrix; (b) load capacity information of dispatched vehicles of the delivery vehicles; and (c) a predetermined average load threshold of the dispatched vehicles. Moreover, the method can include further reassigning, in at least 2 passes, the nodes to the clusters based, at least in part, on the time window information of the nodes. Other embodiments are disclosed.

Automatically planning delivery routes using clustering

A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, perform certain acts. The acts can include obtaining an original load design for loading a trailer to deliver stacks of pallets to physical stores. The load design specifies a respective floor spot assignment for each of the stacks. The original load design satisfies center-of-gravity constraints. The acts also can include generating fluid groups of the stacks to form an initial fluid grouping. The acts additionally can include determining that the fluid groups do not collectively satisfy the center-of-gravity constraints. The acts further can include until the fluid groups collectively satisfy the center-of-gravity constraints, repeatedly determining one or more infeasible groups from the fluid groups, and adjusting at least one of the one or more infeasible groups to update the fluid groups. The acts additionally can include outputting a flexible load design comprising the fluid groups that collectively satisfy the center-of-gravity constraints. Other embodiments are described.

In some examples, a system may include a first computing device communicatively coupled to a second computing device. Additionally, the first computing device is configured to obtain, from the second computing device, check-in data indicating an arrival of the user of the second computing device at a first location, and in response to obtaining the check-in data, determine current wait times. Moreover, the first computing device is configured to determine a first number of customers waiting for service, determine a first number of associates available to assist the first number of customers, and determine an expected wait time for the user operating the second computing device based at least on the current wait times. In some examples, the first number of customers waiting for service, and the first number of associates available. Further, the first computing device is configured to transmit the expected wait time to the second computing device.

Methods and apparatus for determining wait times

Horizontally-scalable systems and methods for scheduling and optimizing deliveries are described herein. At least one scheduler is configured to receive a request to schedule a delivery for an origination location. The request includes a desired time slot. The request is compared to a persistent delivery snapshot for the origination location to determine availability of the desired time slot. An interim delivery snapshot including the requested delivery is generated when the persistent delivery snapshot has an available time slot corresponding to the desired time slot. At least one optimizer is configured to receive the interim delivery snapshot and generate an updated persistent delivery snapshot by applying an optimization process to the interim delivery snapshot.

Growing community of inventors

Palo Alto, CA, United States of America

Mingang Fu