Field of the Invention

[1] This invention relates generally to electronic delivery network scheduling and tracking control systems. More particularly, this invention relates to an intelligently worktime predictive electronic delivery network scheduling and tracking control system which is able to more accurately predicting delivery times for dispatch routing optimisation.

Background of the Invention

[2] Electronic delivery network systems are used to schedule and track deliveries. For example, US 2015/0161563 Al (C OWDSHIPPING, INC.) 11 June 2015 [hereinafter referred to as Dl] discloses an electronic delivery network scheduling and tracking control system which generates travel routes based on spatiotemporal travel route intersection information.

[3] Whereas the system of Dl is able to predict future travel routes and locations, Dl is unable to accurately predict delivery times which is desirous for dispatch routing optimisation.

[4] US 2017/0154347 Al (SIMPLER POSTAGE, INC.) 01 June 2017 [hereinafter referred to as D2] relates to a system for more accurately estimating transit time (and therefore delivery time) for a given shipment which uses cross-carrier delivery prediction models.

[5] The present invention seeks to provide an electronic delivery network scheduling and tracking control system for more accurately predicting delivery times for dispatch routing optimisation, which will overcome or substantially ameliorate at least some of the deficiencies of the prior art, or to at least provide an alternative.

[6] It is to be understood that, if any prior art information is referred to herein, such reference does not constitute an admission that the information forms part of the common general knowledge in the art, in Australia or any other country.

Summary of the Disclosure

[7] There is provided herein an electronic delivery network scheduling and tracking control system configured for more accurately predicting delivery times for shipments for dispatch routing optimisation.

[8] With reference to Figure 1 there is provided an exemplary shipment from a source to a destination, typically using a road network. The total delivery work time comprises a sum of the road network transit time and endpoint handling times which may include source and destination handling times. [9] The present system is directed to more accurately predicting the total delivery time, referred to herein as the delivery work time, for a plurality of future shipments for dispatch routing optimisation including driving efficiencies in job scheduling, clustering and carrier allocation.

[10] The present system may use a GIS-based traffic routing and travel time prediction server for predicting transit times for travel routes using source and destination address data on the basis of prior transit times and real-time traffic conditions.

[11] The present system further uses a supervised machine learning (SM L) subsystem to predict source and/or destination handling times to therefore calculate the total predicted delivery work time.

[12] The supervised machine learning subsystem uses a machine learning module which is trained using previous shipment handling times and associated destination classification data. The trained machine is then able to output a predicted handling times in accordance with at least one destination classification.

[13] The supervised machine learning subsystem may take the form of an artificial neural network (ANN).

[14] The predicted handling time from the machine learning trained machine and the predicted transit time from the traffic routing and travel time prediction server is then used to calculate the predicted delivery work time for the shipment.

[15] The predicted delivery work times are then used by allocation and dispatch control to cluster shipments and allocate parcels to carriers.

[16] In embodiments, the present system employs real-time data from the electronic tracking devices to more accurately measure handling times which are then used by the machine learning module to train a trained machine. These more accurately measured handling times enhance the accuracy of the supervised machine learning subsystem when calculating the predicted handling time.

[17] Specifically, the present system uses carrier electronic tracking devices having GPS receivers for real-time carrier location tracking thereof. In embodiments, the system employs destination address geo-fencing to determine the time of intersection of a carrier location with a destination address geo-fence to more accurately measure the destination handling time. Similarly, the system may employ source address geo-fencing to more accurately measure source handling time.

[18] In embodiments, the system receives data from a receiver electronic device such as by way of haptic input or computer readable media (such as barcode) scan data of a computer readable medium for determining the time of delivery and therefore the more accurate measurement of the destination handling time. [19] Furthermore, in embodiments, the system may calculate travel velocity from the real-time location data to detect variations in velocity to determine the time of transition from the road network to destination premises for more accurate calculation of the destination handling time.

[20] In embodiments, the server may receive real-time accelerometer data from accelerometers of the carrier electronic tracking devices and classify the accelerometer data according to accelerometer classifications to determine the time of transition from the road network to the destination premises, such as, for example, when the carrier transitions from a sedentary travel position to a walk, to more accurately calculate the destination handling time.

[21] In embodiments, the system may further comprise a further supervised machine learning subsystem to automate the classification of destinations. Such a supervised machine learning subsystem similarly uses a machine learning module having as input training data in at least prior destination classification data. The trained machine is optimised by the machine learning module to automatically classify destinations.

[22] In one embodiment, the training data further comprises handling time data associated with the prior destination classification data. In this way, the trained machine is able to accept as input handling time data so as to be able to automate the classification of a destination accordingly.

[23] In a further embodiment, the training data further comprises sensor data received from the carrier electronic tracking devices, such as GPS location and accelerometer motion data. As such, the trained machine is able to accept sensor data and classify a destination automatically accordingly.

[24] In accordance with one aspect, there is provided an electronic delivery network scheduling and tracking control system comprising: a server in operable communication with a plurality of carrier electronic tracking devices across a computer data network, each carrier electronic tracking device comprising a GPS sensor and wherein the server is configured for receiving real-time location data from the plurality of carrier electronic tracking devices accordingly and wherein the server is configured for, for a plurality of shipments each between a source address and a destination address, generating a predicted transit time for an optimised delivery route for each of the plurality of shipments and wherein each destination has at least one destination classification and wherein the server further comprises a supervised machine learning subsystem having a trained machine, the trained machine having been optimised using a machine learning module having as input training data comprising at least one of prior shipment handling time data and associated destination classification data, wherein, for each shipment, the server is configured for using the trained machine to calculate a predicted handling time according to respective at least one destination classification thereof and wherein the server is configured for calculating a predicted work time for each of the shipments in accordance with a respective predicted transit time and a respective predicted handling time and wherein the server is configured for scheduling the shipments at least according to the predicted work time.

[25] The supervised machine learning module may comprise a neural network having at least one input node for the destination classification data and at least one output node for the predicted handling time.

[26] The server may be configured for measuring a prior shipment handling time using sensor data received from at least one respective carrier electronic tracking device and wherein the prior shipment handling time may be used by the machine learning module to train the trained machine.

[27] The sensor data may comprise GPS sensor location data indicative of a real-time location of the at least one respective carrier electronic tracking device and wherein the server may be configured for detecting an intersection of the real-time location and a destination address geo-fence to measure the prior shipment handling time.

[28] The sensor data may comprise GPS sensor location data indicative of a real-time location of the at least one respective carrier electronic tracking device and wherein the server may be configured for calculating the prior shipment handling time in accordance with travel speeds calculated using the real-time location.

[29] The sensor data may comprise accelerometer data indicative of motion of the at least one respective carrier electronic tracking device and wherein the server may be configured for classifying the accelerometer data and calculating the prior shipment handling time according to the accelerometer data classification.

[30] The sensor data may comprise computer readable media scan data and wherein the server may be configured for calculating the prior shipment handling time according to a scan time thereof.

[31] The system may further comprise a further supervised machine learning subsystem comprising a trained machine having been trained using a machine learning module having as input training data comprising at least prior destination classification data and wherein the trained machine may be configured for calculating a destination classification for a destination.

[32] The training data may further comprise prior handling time data associated with the prior destination classification data and wherein the trained machine has as input handling time data.

[33] The training data may further comprise prior sensor data associated with the prior destination classification data and wherein the trained machine has as input sensor data.

[34] Scheduling the shipments may comprise clustering a plurality of shipments according to the according to the predicted work time.

[35] Scheduling the shipments may comprise clustering the relative shipments by carrier.

[36] Other aspects of the invention are also disclosed. Brief Description of the Drawings

[37] Notwithstanding any other forms which may fall within the scope of the present invention, preferred embodiments of the disclosure will now be described, by way of example only, with reference to the accompanying drawings in which:

[38] Figure 1 illustrates an exemplary shipment comprising transit and handling times;

[39] Figure 2 illustrates an electronic delivery network scheduling and tracking control system in accordance with an embodiment;

[40] Figure 3 illustrates supervised machine learning to predict a handling times in accordance with destination classifications in accordance with an embodiment; and

[41] Figure 4 illustrates supervised machine learning to classify destinations according to electronic tracking device sensor data.

Description of Embodiments

[42] Figure 1 illustrates an exemplary shipment of an object, such as a parcel or the like, by a carrier, from a source address 101 to a destination addresses 103 across a road network 102 by vehicular transport.

[43] The total time taken for the shipment 100 from the source 101 to the destination 103 is referred to h erein as the delivery work time 107.

[44] The delivery work time 107 comprises vehicular transit time 105 on the road network 102 and endpoint handling time which may comprise source handling time 104 and destination handling time 106. For example, delivering a parcel may comprise source handling time 104 wherein the carrier arranges for the pickup of the parcel such as at the depot or other type of source 101, transit time 105 wherein the carrier travels by road to the destination 103 and the destination handling time 106 wherein the carrier arranges for the handover of the parcel.

[45] Figure 2 illustrates an exemplary electronic delivery network scheduling and tracking control system 200.

[46] The system 200 comprises a server 200 in operable communication with at least one carrier electronic tracking device 221 across a computer data network 205, such as the Internet. The server 211 may further be in operable communication with at least one client electronic device 222.

[47] In embodiments, the system 200 may be a crowdsourced delivery network wherein a plurality of carriers elect for and are allocated shipments via their respective carrier electronic tracking devices 221 and the shipments may be received via the one or more client electronic device 222.

[48] The server 211 may be in further communication with a plurality of data sources 203, including by way of API access 202. The server 201 may be in operable communication with a GIS-based realtime traffic routing and travel time prediction server 201. The travel time production server 201 receives source and destination GIS address data via the API 202, calculates an optimised route between the addresses and calculates a predicted transit time 105 of the optimised route based on historical traffic data recorded by the server 201, real-time traffic data and other relevant information.

[49] Each of the server 211, carrier electronic tracking device 221 and client electronic device 222 comprises a processor 200 for processing digital data. In operable communication with the processor across a system bus 223 is a memory device 208. The memory device 208 is configured for storing digital data, including computer program code instructions and associated data. In use, the processor 204 fetches these computer program code instructions and associated data from the memory device 208 across a system bus 223 for interpretation and execution for the implementation of the computational functionality described herein. Each device 211, 221 and 222 may further comprise a network interface 212 for sending and receiving data across the data network 205.

[50] For illustrative convenience, the computer program code instructions and associated data have been shown as having been logically divided, including into various controllers as will be described in further detail hereunder.

[51] With reference to the carrier electronic tracking device 221, the device 221 may comprise a plurality of sensors 215, including a GPS receiver 213 and, in embodiments, an accelerometer 214.

[52] The memory device 211 of the carrier electronic tracking device 221 may comprise an application controller 216. The carrier electronic tracking device 221 may further comprise an electronic display device 217 for the display of digital data thereon. In embodiments, a haptic overlay user interface sensor may overlay the electronic device for the receipt of haptic user interface gestures with reference to the electronic data display thereon.

[53] The electronic display device 270 displays a graphical user interface 218 controlled by the application controller 216 which may be used for display of a map representation, including for routing guidance purposes, receipt of input data and the like.

[54] The client electronic device 222 may further comprise an application controller 216 and electronic display device 217 displaying a user interface 218 controlled by the application controller 216.

[55] The server 211 may implement a supervised machine learning (SML) subsystem having a machine learning controller 206 configured for optimising a trained machine 224.

[56] The server 201 may further comprise an allocation and dispatch controller 209 for allocating and dispatching shipments 100. Furthermore, the server 211 may comprise a real-time tracking controller 210 configured for real-time tracking of the carrier electronic tracking devices 221.

[57] In use, the server 201 may receive a plurality of shipments 100. Each shipment comprises source and destination addresses. [58] For each shipment, the server 211 uses the traffic routing and travel time prediction server 201 to calculate the predicted transit time 105 for each shipment 100. For example, the server 211 may send the source and destination address data via the API 202 to the travel time prediction server 201 so as to receive a predicted travel time 105 in reply. In embodiments, the server 201 may also calculate an optimised route.

[59] The server 211 furthermore predicts a handling time which may include the destination handling time 106 and/or the source handling time 104.

[60] Each destination 103 may be classified in accordance with at least one classification. In one embodiment, each destination 103 may be classified according to destination type, such as residential or business destination types for example. The type classification may further comprise sub- classifications wherein, for example, the business type classification may further comprise sub- classifications such as hospital, university et cetera.

[61] It should be noted that, in embodiments, the system 200 need not necessarily use conventional classification such as business, residential and the like and, in embodiments, may use other classifications determinative or representative of destination handling time 106. For example, the system 200 may maintain and, in embodiments, generate classifications including, for example, Classification 1, Classification 2... Classification 3 et cetera.

[62] With reference to Figure 4, there is shown the SML subsystem 300 comprising the server 211 based machine learning module 206 and trained machine 224.

[63] The machine learning module 206 has as input training data 307 which may include prior shipment transit and/or handling time 305 and associated destination classification data 304. The machine learning module 206 generates optimising parameters 319 according to the example input/output pair training data 307 which optimise the trained machine 224.

[64] In embodiments, the SM L machine may comprise a neural network.

[65] As such, for each shipment 100, the destination 103 may comprise a destination classification. As such, the trained machine 224 takes as input real-time data 306 which may comprise destination classification data 303 and outputs a predicted destination handling time 315. In the embodiment wherein the trained machine 224 is a neural network, the neural network may have at least one input note for the destination classification data 303 and at least one output for the predicted handling time 315.

[66] Associated data may also be used to train the trained machine 224 for the calculation of the predicted handling time 315. Associated data may include any type of data which could have an effect on source and/or destination handling times including, for example, time of day information, whether information or the like. Where the trained machine 224 takes the form of a neural network, the trained machine 224 may comprise further input nodes for this associated data.

[67] By using the trained machine 224, the system 200 is able to more accurately calculate the destination handling time 106 and, in embodiments, the source handling time 104.

[68] For example, the supervised machine learning subsystem 300 may learn from historical handling time data that:

a. Residential handling time is typically 23 minutes whereas business handling time is typically 27 minutes.

b. Business handling time in region A is typically 27 minutes whereas business handling time in region B is typically 28 minutes.

c. University handling time is typically 23 minutes between 10 and 11 in the morning but typically 42 minutes between 11AM and 1 PM on Tuesdays.

d. Dentist handling time in region A on Wednesdays between 11 AM and 3 PM is typically 11 minutes unless it is raining in which case it becomes 8 minutes.

[69] These variations in handling time are typically unintuitive and beyond human comprehension and the utilisation of the present supervised machine learning subsystem 300 allows the system 200 to automatically identify unintuitive data covariances which affect handling time which may be used for more accurate calculation of delivery work time.

[70] With reference to Figure 4, the server 211 combines the estimated travel time 311 from the travel time prediction server 201 and the predicted handling time 315 from the trained machine 124 to calculate a predicted work time 312 for each shipment 100.

[71] Shipment clustering control 313 may then be performed on various data, including the predicted work time 312 for each shipment 100. For example, the shipments 100 may be clustered according to carrier availability, carrier location, source and destination locations, vehicle types, vehicle capacities and the like. Usually, shipment clustering control 313 may comprise allocating a plurality of delivery routes to each carrier for the allocated shipments 100. The allocation of the number and type of delivery routes may be more accurately allocated according to the more accurately predicted work time 312.

[72] In embodiments, clustering may be controlled by user configurable parameters including those which seek to optimise the system 200 by minimising the number of delivery runs and those which seek to optimise the system 200 by reducing delivery times.

[73] Allocation control 314 may then be performed wherein push notifications or the like are transmitted to respective carrier electronic tracking devices 221 which may be displayed by the respective user interfaces 218 thereof and acceptance of the shipment allocation responded thereto accordingly. Dispatch and tracking control 316 may then control the dispatch of parcels including the real-time tracking of the carrier electronic tracking devices 221.

[74] In embodiments, the real-time tracking controller 211 of the server 211 is configured for receiving real-time data, including location data, from the plurality of carrier electronic tracking devices 211. In embodiments, this real-time carrier electronic tracking device location data may be used to accurately measure the destination handling time 106.

[75] More accurately measured destination handling time 106 may then be used to enhance the accuracy of the supervised machine learning subsystem 300.

[76] In embodiments, the server 211 may compare a real-time carrier electronic tracking device location with a destination address geo-fence. At the time of the intersection of the real-time carrier electronic tracking device location with the geo-fence, the server 211 may measure the end timestamp of the destination handling so as to more accurately calculate the destination handling time 106. The more accurately calculateed destination handling time 106 in this manner may then be fed back for use as the prior shipment handling times 305 which, in turn, allows the machine learning module 260 train a trained machine 224 more accurately.

[77] In embodiments, the server 211 is further configured for using the real-time carrier electronic tracking device location to more accurately calculate the start timestamp for the destination handling 106 so as to be able to more accurately calculate the estimation handling time 106.

[78] In embodiments, the server 211 may use the real-time carrier electronic tracking device location to calculate a travel speed of the carrier. Differences in measured travel speed may be used to determine the transition from the transit time 105 to the destination handling time 106. For example, the server 211 may calculate an average travel speed for a carrier using a moving window. On the road network 102, the average speed may, for example, be above 40 km/h. However, once having parked the vehicle, dismounted from a motorcycle or the like, the average speed may drop below 12 km/h, indicative of walking. The server 211 may therefore use the real-time carrier electronic tracking device to more accurately calculate the transition from the transit time 105 to the destination handling time 106 which, may in turn, provide more accurate prior shipment transit/handling times 305 for more accurately training the machine learning module 206.

[79] In embodiments, the shipment receiver may use an image sensor 219 of a client electronic device 222 to scan a computer readable media of the parcel, such as a 2D barcode or the like. Scan data may be transmitted in substantial real time to the server 211 which may record a timestamp. This scan time timestamp may be used by the server 211 for more accurately measuring the destination handling time 106 which, in turn may be used for more accurately training the supervised machine learning subsystem 300. [80] In embodiments, the system 200 may use an accelerometer sensor 214 of the carrier electronic tracking device 211 to more accurately calculate the destination handling time 106. For example, the server 211 may compare accelerometer data received in substantial real-time from a carrier electronic tracking device 211 to infer when the carrier is walking. Such comparison may comprise classifying the accelerometer data according to gait indicative accelerometer data. In this way, for example, using accelerometer data, the server 211 is able to detect when the carrier transitions from a more sedentary seated travel position such as in or on a vehicle to walking on premises at the destination.

[81] In embodiments, the system 200 is configured for automated classification of the destinations 103. Specifically, with reference to Figure 4, there is shown a further supervised machine learning subsystem 400 having the machine learning module 206 which optimises the trained machine 406.

[82] Machine learning module 206 is trained using training data 404 which may comprise prior destination classification data 403.

[83] In embodiments, the machine learning module 206 may be trained using prior associated handling time data 402. For example, for a new destination, the system 200 may measure the associated handling time and then classify the type of destination accordingly in an automated manner. For example, the system 200 may determine that a new destination is "University-like", "dentist-like" or the like without requiring human classification thereof.

[84] In further embodiments, the machine learning module 206 may be trained using prior associated sensor data 411. For example, for a new destination, the system 200 may measure sensor data of the carrier electronic tracking devices 211 which, for example, may comprise the GPS location data, accelerometer data or the like. For example, the system 200 may determine from the accelerometer data that the carrier for a destination is walking for approximately 24 minutes, is required ascent several flights of stairs (as determined by the GPS data or via electronic barometric sensor data) and sometimes loses GPS connectivity when walking within multiple buildings and therefore the destination is "campus-like" and therefore the destination should be classified as "university-like".

[85] In embodiments, the destination classifications may be supplemented with destination categorisation parameter data which may affect the destination handling time. For example, a destination classification may be a "business" classification and wherein the associated parameter data may include parameter data such as "has stairs", "has elevators" or the like which may, for example, affect the carrier reaching the drop-off location in a multistorey building. In embodiments, these parameters may be automatically determined by the system 200 in accordance with sensor data, such as location and accelerometer sensor data received from the carrier electronic devices. [86] The foregoing description, for purposes of explanation, used specific nomenclature to provide a thorough understanding of the invention. However, it will be apparent to one skilled in the art that specific details are not required in order to practice the invention. Thus, the foregoing descriptions of specific embodiments of the invention are presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise forms disclosed; obviously, many modifications and variations are possible in view of the above teachings. The embodiments were chosen and described in order to best explain the principles of the invention and its practical applications, they thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. It is intended that the following claims and their equivalents define the scope of the invention.