Abstract: Provided are systems and processes for optimizing assignments of deliveries for perishable goods. In one aspect, a method is provided for pairing a set of created orders with a set of available couriers. The set of created orders may include orders confirmed by the merchant and the set of available couriers include couriers that are online with an active status. Feasible pairings are generated between each courier and each created order. Infeasible pairings are eliminated based on factors such as transportation mode. Possible routes for each pairing are generated and scored based on weighted factors. The scores are optimized to achieve a set of routes with a maximum score. The routes are then offered to the corresponding courier if the courier will arrive at or after the created order is completed by the merchant. A neural network may be implemented to recognize the optimal score for a given duration.

Abstract: In some examples, a service computing device may receive, from respective mobile devices of a plurality of mobile devices, an indication of a communication availability condition of each respective mobile device. The service computing device may determine, for a delivery job, based at least in part on the indication of the communication availability condition received from each of the respective mobile devices, that a first mobile device of the plurality of mobile devices is likely to have a greater communication availability for a duration of the delivery job than a second mobile device of the plurality of mobile devices. Based at least partially on the determination, the service computing device may send, to the first mobile device, a communication to assign the delivery job to a courier associated with the first mobile device.

Abstract: Provided are systems and processes for optimizing assignments of deliveries for perishable goods. In one aspect, a method is provided for pairing a set of created orders with a set of available couriers. The set of created orders may include orders confirmed by the merchant and the set of available couriers include couriers that are online with an active status. Feasible pairings are generated between each courier and each created order. Infeasible pairings are eliminated based on factors such as transportation mode. Possible routes for each pairing are generated and scored based on weighted factors. The scores are optimized to achieve a set of routes with a maximum score. The routes are then offered to the corresponding courier if the courier will arrive at or after the created order is completed by the merchant. A neural network may be implemented to recognize the optimal score for a given duration.

Abstract: In some examples, a service provider may receive an indicated delivery location from a buyer device. The service provider may determine, e.g., based in part on traffic information, respective predicted courier travel times to the delivery location from one or more different merchant locations. The service provider may compare the respective predicted courier travel times with respective spoilage times for items offered by the corresponding one or more merchants, and may identify one or more items having respective spoilage times greater than the respective predicted courier travel time from that merchant location. The service provider may send, to the buyer device, information about items available for delivery from particular merchants having respective spoilage times greater than the predicted courier travel time from respective the merchant location. The buyer device may present the information about the items on a display to enable buyer selection of an item for delivery.

Abstract: Provided are various mechanisms and processes for generating dynamic merchant scoring predictions. A system is configured to receive datasets comprising pairings between training customer profiles and training merchant profiles. For each pairing, a set of feature values corresponding to features specified by the customer and merchant profiles are extracted and converted into a training vector to train a machine learning model to determine a weighted coefficient for each feature. Once sufficiently trained, the system determines a set of available merchant profiles for a customer profile in response to receiving a search request from a customer associated with the customer profile. For each pairing between the customer profile and an available merchant profile, the system determines an order score for the available merchant based on the weighted coefficients and an input set of feature values specified by the customer profile and the available merchant profile.

Abstract: In some examples, a computing device may receive, from a first user device, a communication indicating a request to create a combined order for items offered by one or more merchants. For example, the combined order may enable a plurality of users to independently purchase individual ones of the items from the one or more merchants. The combined order may be created having an initial condition and may be accessible by the plurality of users for adding items to the combined order. At least one selection of a respective item may be received from each user of the plurality of users. For instance, each selection may change the condition of the combined order. In addition, the computing device may process the items added to the combined order as a single order to perform at least one action.

Abstract: Provided are various systems and processes for improving last-mile delivery of real-time, on-demand orders for perishable goods. In one aspect, an automated vehicle (AV) comprises a body including a storage compartment for storing perishable goods. The storage compartment is accessible by a user upon authentication of the user. The AV further comprises a sensor module for receiving data for navigating the AV. The sensor module is positioned above the body on a support structure at a predetermined height above the ground, such as three to five feet. The data includes one or more of the following: audio data, video data, radio waves, and backscattered light waves. The AV further comprises an onboard computer system configured to process the data to navigate the AV along motor vehicle routes and pedestrian routes. The AV may be configured to interface with an automated locker system to retrieve or deposit the perishable goods.

Abstract: Provided are various systems and processes for improving last-mile delivery of real-time, on-demand orders for perishable goods. In one aspect, a method is provided for aggregating on-demand deliveries using a depot dispatch protocol which may implement automated order transport and retrieval systems. The method comprises dispatching merchant couriers to transport on-demand orders from merchants to a merchant depot where the orders are aggregated and batched based on optimized delivery routes and destination proximities. Batches of orders are then transported to a customer depot corresponding to an area of delivery destinations. Orders are then assigned to delivery couriers for completion of delivery to customers. Such delivery routing systems and processes may be implemented alongside a delivery tracking system for generating estimated time of arrival predictive updates for real-time delivery of perishable goods.

Abstract: In some examples, a smart container may enable crowdsourced delivery of items from a pickup location to a delivery location. The container may include a location sensor and a communication interface to enable the container to provide its location to at least one of a service computing device, a courier device, a sender device, or a recipient device. In some cases, a service provider may employ public transport vehicles to transport the container along a portion of a delivery route. One crowdsourced courier may pick up the container for placement on the public transport vehicle and another crowdsourced courier may pick up the container from the public transport vehicle and deliver the container to the delivery location. Further, the container may include multiple compartments, and recipients who receive a shipment in the container may be solicited to deliver the container to a delivery location of a next recipient.

Abstract: A remote vehicle control system includes a vehicle mounted sensor system including a video camera system for producing video data. A data handling system is connected to a network to transmit data to and receive data from a remote teleoperation site. A virtual control system is configured to receive the video, provide a user with a live video stream supported by machine intelligence directed frame synthesis, and transmit control instructions to the remote vehicle over the network. The frame synthesis is supported by a convolutional neural network. The frame synthesis may be used to interpolate frames to increase effective frames per second. The frame synthesis may be used to extrapolate frames to replace missing or damaged video frames.

Abstract: In some examples, a service provider may receive a plurality of communications indicating buyer accesses to respective item-ordering applications on respective buyer devices. The service provider may determine that the buyer accesses occurred at respective buyer locations outside of a current delivery region. The service provider may determine a cluster of the respective buyer locations, and may further determine, based on the cluster, a recommended courier location for extending the delivery region to include a new area including the cluster. Additionally, in some cases, the service provider may determine a recommended courier location based on a plurality of past delivery locations. Further, in response to receiving an indication of a delivery location from a buyer device, the service provider may determine that a courier is within a threshold delivery travel time to the delivery location prior to providing item information to the buyer device.

Abstract: Provided are various mechanisms and processes for generating dynamic merchant similarity predictions. In one aspect, a system is configured for receiving historical datasets that include a series of merchants from historical browsing sessions generated by one or more users. The merchants are converted into corresponding vector representations for training a predictive model to output associated merchants based on a generated weighted vector space. Once sufficiently trained, data from a new browsing session may be received, which may include a target merchant. The target merchant is input into the predictive model as a vector to output one or more context merchants having vectors with the highest cosine similarity value to the target merchant vector. Selected context merchants may then be transmitted to the user device as targeted merchant suggestions in the new browsing session. The predictive models may be continuously trained using data received from subsequent browsing sessions.

Abstract: Described are systems and processes for generating dynamic effort-based delivery value predictions for real-time delivery of perishable goods. In one aspect, a system is configured for generating dynamic delivery value predictions for delivery opportunities provided to couriers. For each order, delivery events and corresponding timestamps are received from devices operated by customers, restaurants, and couriers. Based on the timestamps, the system generates a predicted delivery duration with trained predictive models that use weighted factors such as order data and historical restaurant data. A service value for the delivery of the order is determined based the predicted delivery duration and a predetermined active time value. The service value is then transmitted along with the corresponding delivery opportunity to a user device of a courier. The determined service values may be adjusted based on courier acceptance rates of delivery opportunities and other factors such as customer experience.

Abstract: Provided are systems and processes for optimizing assignments of deliveries for perishable goods. In one aspect, a method is provided for pairing a set of created orders with a set of available couriers. The set of created orders may include orders confirmed by the merchant and the set of available couriers include couriers that are online with an active status. Feasible pairings are generated between each courier and each created order. Infeasible pairings are eliminated based on factors such as transportation mode. Possible routes for each pairing are generated and scored based on weighted factors. The scores are optimized to achieve a set of routes with a maximum score. The routes are then offered to the corresponding courier if the courier will arrive at or after the created order is completed by the merchant. A neural network may be implemented to recognize the optimal score for a given duration.

Abstract: Provided are mechanisms and processes for facilitating a real-time, on-demand delivery platform for perishable goods by determining aspects such as the location and status of a courier who is affiliated with a delivery platform. In one example, a courier carries a Bluetooth Low Energy (BLE) enabled mobile device that is configured to detect signals from beacons located at both a vehicle registered to the courier and a provider site. Each of the detected signals includes identifying information and signal strength of the detected signal. The identifying information and signal strengths are transmitted to a server that monitors the signal strengths received at the courier's BLE-enabled mobile device from the beacon devices to determine the likelihood that the courier has arrived at the provider site.

Abstract: Described are systems and processes for generating dynamic estimated time of arrival predictive updates for delivery of perishable goods. In one aspect a system is configured for generating dynamic estimated time of arrival (ETA) predictive updates between a series of successive events for real-time delivery of orders. For each order, a plurality of delivery events and corresponding timestamps are received from devices operated by customers, restaurants, and couriers. Based on the timestamps, the system generates a plurality of ETA time predictions for one or more of the delivery events with trained predictive models that use weighted factors including historical restaurant data and historical courier performance. As additional timestamps are received for a delivery event, the trained predictive models dynamically update the ETA time predictions for successive events. The predictive models may be continuously trained by updating the weighted factors based on the received timestamps.

Abstract: Provided are various mechanisms and processes for optimizing real-time, on-demand deliveries of perishable goods. In one aspect, a method is provided for aggregating on-demand deliveries using a depot dispatch protocol. The method comprises dispatching runners to pick up on-demand orders for drop off at a depot where the orders are aggregated and batched based on optimized delivery routes and destination proximities. Batches of orders are then assigned to couriers who may arrive at the depot to receive the batched orders without having to navigate through areas of high congestion. Such delivery routing system may be implemented alongside a delivery tracking system for generating estimated time of arrival predictive updates for real-time delivery of perishable goods.