Fill modeling for hybrid last-mile delivery

1. A system of filling a delivery vehicle for hybrid last-mile delivery, comprising:

a processor programmed to:

access training data comprising data from prior filled delivery vehicles and a plurality of features associated with the prior filled delivery vehicles;

train a machine-learning (ML) model to learn relationships between the plurality of features and labeled data indicating whether a prior fill level of the prior filled delivery vehicles from the training data was an appropriate fill level, the labeled data including at least one of an amount of time for delivery at a pickup location and feedback from at least one prior customer indicating a level of satisfaction of the pickup location;

generate a fill-model using the ML model, the fill-model outputting at least one of a first probability that a fill level of the delivery vehicle is a suitable fill level and a second probability that the fill level of the delivery vehicle is not a suitable fill level;

access an order queue comprising one or more pickup orders to be picked up by respective customers at a pickup location;

identify, based on the fill-model, a plurality of items, ordered by the respective customers and associated with the one or more pickup orders, that satisfy a fill criterion comprising a minimum fill level of the delivery vehicle to service the pickup location, the minimum fill level defining a minimum amount of a capacity of the delivery vehicle to be filled to ensure that the capacity of the delivery vehicle is filled with items at least to the minimum amount;

wherein to identify the plurality of items based on the fill-model, the processor is programmed to:

provide, as input to the ML model, data relating to the plurality of items, the data relating to the plurality of items having respective values of the plurality of features; and

determine, based on an output of the ML model, that the plurality of items satisfies the fill criterion;

responsive to identification of the plurality of items that satisfy the fill criterion, trigger the plurality of items to be loaded onto the delivery vehicle at a distribution center to initiate delivery of the plurality of items;

transmit, to a delivery vehicle device, the one or more pickup orders corresponding to the plurality of items and a specified time at which to be at the pickup location to meet the respective customers; and

transmit, to each customer associated with the plurality of pickup orders, a notification to meet the delivery vehicle to pick up the plurality of items at the pickup location and the specified time at which to meet the delivery vehicle for pickup.

2. The system of claim 1 , wherein the fill-model comprises a rules-based model, and wherein to apply the fill-model, the processor is programmed to:

access one or more fill rules that specify the fill criterion, the fill criterion specifying a minimum number of items, a minimum value of items, a minimum volume of items to consider the delivery vehicle to be filled;

determine whether the plurality of items satisfy the one or more fill rules.

3. The system of claim 2 , wherein the processor is programmed to:

iterate through the one or more orders to identify different sets of items to be evaluated against the one or more fill rules.

4. The system of claim 1 , wherein the ML model is trained to identify items that will likely fit within the delivery vehicle based on prior deliveries.

5. The system of claim 1 , wherein the ML model is trained to identify items that will likely be picked up without cancelations by the respective customers, as determined from prior deliveries from the training data, based on a probability of cancelation, and wherein to determine that the plurality of items satisfies the fill criterion, the processor is further programmed to: identify the plurality of items based on respective probabilities of cancelation.

6. The system of claim 1 , wherein the processor is further programmed to:

prior to the identification the plurality of items that satisfies the fill criterion, determine that a previously evaluated plurality of items, from among the plurality of items, did not satisfy the fill criterion;

identify one or more items that, when added to the previously evaluated plurality of items, will satisfy the fill criterion;

identify one or more customers that may be interested in the identified one or more items; and

transmit, to the identified one or more customers, a recommendation to add the identified one or more items.

7. The system of claim 1 , wherein the plurality of items includes retail items to be delivered to a retail partner at a retail address, and wherein the processor is further programmed to:

set a delivery route for the delivery vehicle, the delivery route including the pickup location and the retail address.

8. The system of claim 1 , wherein the processor is further programmed to:

determine that the delivery vehicle is over-capacity;

identify one or more items that would address the over-capacity; and

transmit an incentive to a customer associated with the one or more items to accept an alternate pickup location or alternative pickup time.

9. The system of claim 1 , wherein the processor is further programmed to:

determine an arrival order in which customers are predicted to arrive at the pickup location to meet the delivery vehicle; and

determine a loading order in which a plurality of pickup orders of the customers is to be loaded into the delivery vehicle based on the determined arrival order.

10. The system of claim 1 , wherein the processor is further programmed to:

determine, for each customer, a probability that the customer will pick up a respective pickup order at the pickup location; and

determine a loading order in which a plurality of pickup orders of the customers is to be loaded into the delivery vehicle based on the determined probability determined for each customer.

11. The system of claim 1 , wherein the delivery vehicle is to deliver to the pickup location with a first plurality of pickup orders and a second location with a second plurality of pickup orders, and wherein the processor is further programmed to:

determine a first loading order of the first plurality of pickup orders in a first section of the delivery vehicle; and

determine a second loading order of the second plurality of pickup orders in a second section of the delivery vehicle.

12. A method of filling a delivery vehicle for hybrid last-mile delivery, comprising:

accessing training data comprising data from prior filled delivery vehicles and a plurality of features associated with the prior filled delivery vehicles;

training a machine-learning (ML) model to learn relationships between the plurality of features and labeled data indicating whether or not a prior fill level of the prior filled delivery vehicles from the training data was an appropriate fill level, the labeled data including at least one of an amount of time for delivery at a pickup location and feedback from at least one prior customer indicating a level of satisfaction of the pickup location;

generating a fill-model using on the ML model, the fill-model outputting at least one of a first probability that a fill level of the delivery vehicle is a suitable fill level and a second probability that the fill level of the delivery vehicle is not a suitable fill level;

accessing, by a processor, an order queue comprising one or more pickup orders to be picked up by respective customers at a pickup location;

identifying, by the processor, based on the fill-model, a plurality of items, ordered by the respective customers and associated with the one or more pickup orders, that satisfy a fill criterion comprising a minimum fill level of the delivery vehicle to service the pickup location, the minimum fill level defining a minimum amount of a capacity of the delivery vehicle to be filled to ensure that the capacity of the delivery vehicle is filled with items at least to the minimum amount,

wherein identifying the plurality of items based on the fill-model comprises:

providing, as input to the ML model, data relating to the plurality of items, the data relating to the plurality of items having respective values of the plurality of features; and

determining, based on an output of the ML model, that the plurality of items satisfies the fill criterion;

wherein the plurality of items includes retail items to be delivered to a retail partner at a retail address;

responsive to identification of the plurality of items that satisfy the fill criterion, triggering, by the processor, the plurality of items to be loaded onto the delivery vehicle at a distribution center to initiate delivery of the plurality of items;

transmitting, by the processor, to a delivery vehicle device, the one or more pickup orders corresponding to the plurality of items and a specified time at which to be at the pickup location to meet the respective customers;

transmitting, by the processor, to each customer associated with the plurality of pickup orders, a notification to meet the delivery vehicle to pick up the plurality of items at the pickup location and the specified time at which to meet the delivery vehicle for pickup; and

setting, by the processor, a delivery route for the delivery vehicle, the delivery route including the pickup location and the retail address.

13. The method of claim 12 , wherein the fill-model comprises a rules-based model, and wherein applying the fill-model comprises:

accessing one or more fill rules that specify the fill criterion, the fill criterion specifying a minimum number of items, a minimum value of items, a minimum volume of items to consider the delivery vehicle to be filled; and

determining whether the plurality of items satisfy the one or more fill rules.

14. The method of claim 13 , further comprising:

iterating through the one or more orders to identify different sets of items to be evaluated against the one or more fill rules.

15. The method of claim 12 , wherein the ML model is trained to identify items that will likely fit within the delivery vehicle based on prior deliveries from the training data, based on a probability of cancelation, and wherein identifying the plurality of items that satisfies the fill criterion comprises identifying the plurality of items based on respective probabilities of cancelation.

16. The method of claim 12 , wherein the ML model is trained to identify items that will likely be picked up without cancelations based on prior deliveries.

17. A non-transitory computer readable medium storing instructions that, when executed by a processor, causes the processor to:

access training data comprising data from prior filled delivery vehicles and a plurality of features associated with the prior filled delivery vehicles;

train a machine-learning (ML) model to learn relationships between the plurality of features and labeled data indicating whether or not a prior fill level of the prior filled delivery vehicles from the training data was an appropriate fill level, the labeled data including at least one of an amount of time for delivery at a pickup location and feedback from at least one prior customer indicating a level of satisfaction of the pickup location,

access an order queue comprising one or more pickup orders to be picked up by respective customers at a pickup location;

generate a fill-model using on the ML model, the fill-model outputting at least one of a first probability that a fill level of the delivery vehicle is a suitable fill level and a second probability that the fill level of the delivery vehicle is not a suitable fill level;

identify, based on the fill-model, a plurality of items, ordered by the respective customers and associated with the one or more pickup orders, that satisfy a fill criterion comprising a minimum fill level of a delivery vehicle to service the pickup location, the minimum fill level defining a minimum amount of a capacity of the delivery vehicle to be filled to ensure that the capacity of the delivery vehicle is filled with items at least to the minimum amount,

wherein to identify the plurality of items based on the fill-model, the instructions, when executed by the processor, further cause the processor to:

provide, as input to the ML model, data relating to the plurality of items, the data relating to the plurality of items having respective values of the plurality of features; and

determine, based on an output of the ML model, that the plurality of items satisfies the fill criterion;

responsive to identification of the plurality of items that satisfy the fill criterion, trigger the plurality of items to be loaded onto the delivery vehicle at a distribution center to initiate delivery of the plurality of items;

transmit, to a delivery vehicle device, the one or more pickup orders corresponding to the plurality of items and a specified time at which to be at the pickup location to meet the respective customers; and

transmit, to each customer associated with the plurality of pickup orders, a notification to meet the delivery vehicle to pick up the plurality of items at the pickup location and the specified time at which to meet the delivery vehicle for pickup.

18. The system of claim 1 , wherein the processor is further programmed to:

after the identification of the plurality of items that satisfy the fill criterion, receive an indication that an item in the plurality of items has been canceled from a pickup order;

identify one or more customers that may be interested in the canceled item or a replacement item instead of the canceled item that will fulfill the fill criterion; and

transmit, to the identified one or more customers, a recommendation to add the canceled item or the replacement item.