Parcel handling for electronic parcel compartment systems

Abstract

The present disclosure relates to the general field of logistics and, more specifically, last-mile deliveries. In more detail, the present disclosure relates to the handling of parcels, e.g. parcels that are to be delivered to an electronic parcel compartment system, where the compartments of the electronic parcel compartment system have different sizes to accommodate parcels of different sizes. In one aspect described herein, a computer-implemented method (700) implemented in an apparatus for matching of a parcel with dimensionally-constrained lockers of an electronic parcel compartment system is proposed.

Description

[0001] PARCEL HANDLING FOR ELECTRONIC PARCEL COMPARTMENT SYSTEMS

[0002] Technical Field

[0003] The present disclosure relates to the general field of logistics and, more specifically, last-mile deliveries. In more detail, the present disclosure relates to the handling of parcels, e.g. parcels which are to be delivered to an electronic parcel compartment system, where the compartments of the electronic parcel compartment system have different sizes to accommodate parcels of different sizes.

[0004] The present disclosure proposes a method and a corresponding apparatus for generating a dataset for a machine-learning algorithm. Furthermore, a computer- implemented method and corresponding apparatus for training a machine-learning algorithm to model the matching of parcels with dimensionally-constrained lockers of an electronic parcel compartment system are proposed. Still further, a computer- implemented method and a corresponding apparatus for matching of a parcel with dimensionally-constrained lockers of an electronic parcel compartment system are proposed.

[0005] Background

[0006] Locker systems are known in the art. Generally, a locker system comprises a plurality of compartments. Each compartment comprises a respective door that, in an opened state, allows objects to be placed into or retrieved from an interior of the compartment. A lock associated with the door allows the door to be locked (nonopenable) or unlocked (openable) and thus allows for controlling access to the compartment. It is common that the locks are operated electronically. Therefore, locker systems are often called electronic parcel compartment system.

[0007] It is known in the art that the different compartments of the locker system come with different sizes (volumes). Accordingly, the compartments are sized to accommodate parcels having different spatial dimensions. This has the advantage that courier service companies can offer last-mile deliveries of various kinds of parcels having different spatial dimensions to the locker system, where the users either retrieve (pick up) or return the parcels.

[0008] One of the challenges with deliveries to destinations such as electronic parcel compartment systems is the fact that there is an uncertainty about the available suitably sized lockers, or compartments, of the electronic parcel compartment system. For example, in the past when a courier was unable to deliver a parcel to a particular address (e.g., because the parcel did not fit into any of the available lockers of the electronic compartment system), the courier would typically attempt to redeliver the parcel on another day. Such redelivery attempts could occur multiple times for a particular parcel, which would result in an increased number of travels for the courier. A high number of travels may have a negative impact on the environmental footprint. Also, this may have the negative effect of increased transactions costs for the courier and, obviously, unwelcome delays in receiving the parcels for the recipient.

[0009] Summary

[0010] It is in view of the above considerations and others that the various aspects and embodiments of the present invention have been made.

[0011] In particular, the inventors have realized that there is a need for an improved technique for facilitating delivery of parcels to suitably sized lockers of an electronic parcel compartment system. Furthermore, the inventors have realized that it may be a challenge to know in advance which parcels that fit into which compartments of an electronic parcel compartments system. In particular, this challenge may be particularly noticeable for flexible parcels, such as bags. In view of this, the present disclosure therefore recognizes the fact that there is a need for alternatives to (e.g. improvement of) the existing art described above. It is an object of some aspects and embodiments described herein to solve, mitigate, alleviate, or eliminate one or more of the disadvantages with the prior art described above.

[0012] A general object of the present disclosure is therefore to provide methods, apparatuses and computer program that allow for a facilitated delivery of parcels of different shapes and sizes to suitably sized lockers of an electronic parcel compartment system. According to a first aspect, a method of generating a dataset for a machinelearning algorithm is proposed. As will be appreciated by those skilled in the art, a purpose of this aspect is to provide a unique process of generating this dataset, which can then be used as the training dataset for the machine-learning algorithm. The method comprises: obtaining a second plurality of parcels, which is a subset of larger first plurality of parcels; determining, or otherwise sensing, spatial dimensions of each one of the obtained second plurality of parcels; dependent on spatial dimensions of one or more separate dimensionally- constrained lockers of an electronic parcel compartment system, determining in which dimensionally-constrained locker(s) a parcel can be placed; assigning each parcel a value indicative of the smallest possible dimensionally-constrained locker of the electronic parcel compartment system in which it can be placed; grouping the parcels of the obtained second plurality of parcels into different groups of parcels based on the assigned values; and generating, or otherwise forming, a dataset including data relating to the different groups of parcels based on the assigned values.

[0013] In some embodiments, the method comprises the following actions, or method steps, prior to obtaining the second plurality of parcels: scanning said first plurality of parcels to obtain first data relating to spatial dimensions of each one of the parcels in the first plurality of parcels; and

[0014] - based on the first data determining, or otherwise predicting, whether a parcel is either one of two types of parcels; and wherein obtaining the second plurality of parcels includes obtaining only parcels that are determined, or otherwise predicted, to be of one of the two types of parcels.

[0015] In some embodiments, the method further comprises, prior to obtaining the second plurality of parcels: obtaining second data relating to a source, such as a merchant, associated with the first plurality of parcels; and

[0016] - based on a combination of the first and second data determining, or otherwise predicting, whether a parcel is either one of two types of parcels; and wherein obtaining the second plurality of parcels includes obtaining only parcels that are determined, or otherwise predicted, to be of one of the two types of parcels.

[0017] For example, a first type of parcel may be an inflexible parcel (such as a box) and a second type of parcel may be a flexible parcel (such as a bag). In such embodiments, the action or method step of obtaining the second plurality of parcels may include obtaining only parcels that are determined, or otherwise predicted, to be of the second type.

[0018] In some embodiments, the action or method step of determining, or otherwise predicting, whether a parcel is either one of two types of parcels may include utilizing a machine-learning algorithm for classifying each parcel of the plurality of parcels into either one of the two types of parcels. The classification of the parcels into either one of the two types of parcels may be performed in various ways. For instance, the method may comprise utilizing one or a combination of i) a K-nearest neighbours (KNN) classification algorithm; ii) a Radius Neighbours classification algorithm, and iii) a support vector machine algorithm. Additionally, or alternatively, the method may comprise utilizing a logistics regression algorithm.

[0019] In some embodiments, the action or method step of determining in which dimensionally-constrained locker(s) a parcel can be placed comprises utilizing a machine-learning algorithm, wherein the machine learning algorithm utilizes one or a combination of i) a KNN classification algorithm, ii) a weighted KNN classification algorithm, and iii) a logistic regression algorithm.

[0020] According to a second aspect, an apparatus comprising means adapted to carry out the method according to the first aspect is provided. To this end, the apparatus may comprise one or more processors and one or more memories. The one or more memories may comprise instructions executable by the one or more processors whereby the apparatus is operative to perform at least some of the actions or method steps of the method according to the first aspect.

[0021] According to a third aspect, a computer program comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out the method according to the first aspect is provided. A carrier comprising the computer program according to the third aspect is also provided. The carrier may be any one of an electronic signal, an optical signal, a radio signal, or a computer-readable medium.

[0022] According to a fourth aspect, a computer-implemented method of training a machine-learning algorithm implemented in an apparatus to model the matching of parcels with dimensionally-constrained lockers of an electronic parcel compartment system is provided.

[0023] The method comprises:

[0024] (a) initialising a machine learning algorithm;

[0025] (b) optionally obtaining a data set including data indicative of different groups of parcels based on assigned values, wherein the assigned values are indicative of a smallest possible dimensionally- constrained locker of the electronic parcel compartment system in which a parcel can be placed;

[0026] (c) providing a plurality of parcels being members of a first group of parcels and thus being capable of being placed in the same smallest possible dimensionally-constrained locker of the electronic parcel compartment system;

[0027] (d) scanning spatial dimensions of the parcels of the plurality of parcels of the first group of parcels;

[0028] (e) repeating (c)-(d) for n groups of parcels until all n groups of parcels have been scanned;

[0029] (f) updating the machine learning algorithm accordingly.

[0030] In advantageous embodiments, the obtained data set may have been generated by a method according to the first aspect described above. In some embodiments, the computer-implemented method is repeated m number of times.

[0031] According to a fifth aspect, an apparatus comprising means adapted to carry out the method according to the fourth aspect is provided. To this end, the apparatus may comprise one or more processors and one or more memories. The one or more memories may comprise instructions executable by the one or more processors whereby the apparatus is operative to perform at least some of the actions or method steps of the method according to the fourth aspect. In some embodiments, the apparatus may be embodied a sorting machine.

[0032] According to a sixth aspect, a computer program comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out the method according to the fourth aspect is provided. A carrier comprising the computer program according to the third aspect is also provided. The carrier may be any one of an electronic signal, an optical signal, a radio signal, or a computer-readable medium.

[0033] According to a seventh aspect, a computer-implemented method implemented in an apparatus for matching of a parcel with dimensionally-constrained lockers of an electronic parcel compartment system is provided. A purpose of this aspect is to allow for improved predictions (e.g. in advance of a delivery) of which parcels that fit into which lockers / compartments. The method comprises: scanning a parcel to determine spatial dimensions of the parcel; applying the trained machine learning algorithm of the fourth aspect to the determined spatial dimensions of the parcel; and based on this producing a value indicative of a probability that the scanned parcel can fit inside one or several of the dimensionally-constrained lockers of the electronic parcel compartment system.

[0034] According to an eighth aspect, an apparatus comprising means adapted to carry out the method according to the fourth aspect is provided. To this end, the apparatus may comprise one or more processors and one or more memories. The one or more memories may comprise instructions executable by the one or more processors whereby the apparatus is operative to perform at least some of the actions or method steps of the method according to the fourth aspect. In some embodiments, the apparatus may be embodied as a sorting machine.

[0035] According to a ninth aspect, a computer program comprising instructions which, when executed on at least one processor, cause the at least one processor to carry out the method according to the fourth aspect is provided. A carrier comprising the computer program according to the third aspect is also provided. The carrier may be any one of an electronic signal, an optical signal, a radio signal, or a computer-readable medium.

[0036] Various aspects and embodiments described herein are advantageous in that they enable an improvement over the existing art. Some aspects and embodiments proposed herein allow for a facilitated delivery of parcels of different shapes and sizes to suitably sized lockers of an electronic parcel compartment system. For example, the inventors have recognized that it is beneficial to know (in advance) a probability that a parcel will fit inside one or more lockers of an electronic parcel compartment system. By knowing this before deliveries it is possible to plan deliveries more accurately and, in turn, this enables a facilitated delivery of parcels of different shapes and sizes to suitably sized lockers of an electronic parcel compartment system. The deliveries, and thus transports, can also be planned more cleverly and fewer transports may be required for delivering a certain amount of parcels. Fewer and smarter transports may also lead to a reduced environmental footprint by delivery services.

[0037] Furthermore, it may be possible to provide adjustable sizes of the compartments of the electronic parcel compartment system. By providing an electronic compartment system with compartments with different sizes, it is generally possible to fit more parcels into the electronic compartment system. This way more parcels could optionally fit into one and a same compartment, since smaller parcels may only take up a minimal space of this compartment. This may contribute even further to fewer deliveries and, hence, a further reduced environmental footprint. Brief Description of the Drawings

[0038] These and other aspects, features and advantages will be apparent and elucidated from the following description of various embodiments, reference being made to the accompanying drawings, in which:

[0039] Fig. 1 illustrates an example of an electronic parcel system;

[0040] Fig. 2 illustrates a flowchart of a method according to a first aspect of the disclosure;

[0041] Fig. 3 illustrates a flowchart of a method according to a second aspect of the disclosure;

[0042] Figs. 4A and B are an illustration and a corresponding photograph of collected flexible parcels (bags) during an experiment made by the applicant;

[0043] Figs. 5-6 illustrate an example user interface at a sorting machine; and

[0044] Fig. 7 illustrates a flowchart of a method according to a second aspect of the disclosure;

[0045] Fig. 8 illustrates an example implementation of an embodiment of data processing system; and

[0046] Fig. 9 illustrates a carrier containing a computer program, in accordance with an embodiment.

[0047] Detailed Description of Embodiments

[0048] The present invention will now be described more fully hereinafter. The invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided by way of example so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those persons skilled in the art. Like reference numbers refer to like elements throughout the description.

[0049] As described above, there is a need for a facilitated delivery of parcels of different shapes and sizes to suitably sized lockers of an electronic parcel compartment system. The following description proposes various methods and corresponding apparatuses to address this need. Reference is made to Fig. 1, which illustrates an example of an electronic parcel compartment system 100. The electronic parcel compartment system 100 may be used for so called last mile deliveries. As can be seen in Fig. 1, the electronic parcel compartment system 100 may comprise a plurality of compartments. Typically, the compartments have different sizes (volumes). That is, the compartments may have different widths, heights, and / or depths. Since the compartments have different sizes (volumes), the compartments are sized to accommodate parcels (or, items) of various sizes and shapes. In addition, certain compartments may optionally accommodate more than one parcel. Each compartment may be equipped with a respective door that, in an opened state, allows objects to be placed into or retrieved from an interior of the compartment. A lock associated with the door allows the door to be locked (nonopenable) or unlocked (openable) and thus allows for controlling access to the compartment.

[0050] As discussed above, one of the challenges with deliveries to destinations such as electronic parcel compartment systems 100 is the fact that there is an uncertainty about the available suitably sized lockers, or compartments, of the electronic parcel compartment system 100. For example, in the past when a courier was unable to deliver a parcel to a particular address (e.g., because the parcel did not fit into any of the available lockers of the electronic compartment system), the courier would typically attempt to redeliver the parcel on another day. Such redelivery attempts could occur multiple times for a particular parcel, which would result in an increased number of travels for the courier. A high number of travels may have a negative impact on the environmental footprint. Also, this may have the negative effect of increased transactions costs for the courier and, obviously, unwelcome delays in receiving the parcels for the recipient.

[0051] In the light of this and other considerations, the inventors have realized that there is a need for an improved technique for facilitating delivery of parcels to suitably sized lockers of an electronic parcel compartment system 100. Furthermore, the inventors have realized that it may be a challenge to know in advance which parcels that fit into which compartments of the electronic parcel compartments system 100. In particular, this disclosure recognizes the fact that this has proven to be particularly noticeable for flexible parcels, such as bags. Flexible parcels, such as bags, typically have a shape and properties which make them formable. These parcels may e.g. be squeezed and folded into smaller sizes. Therefore, flexible parcels such as bags are often foldable and could possibly fit into smaller compartments than initially intended or thought.

[0052] In the following, the present disclosure will present various methods, apparatuses and computer programs that address this. In particular, the methods, apparatuses and computer programs proposed herein aim for enabling a facilitated delivery of parcels of different shapes and sizes to suitably sized lockers of an electronic parcel compartment system 100.

[0053] Fig. 2 illustrates a flow chart of illustrating a method 200 of generating a dataset for a machine-learning algorithm.

[0054] Action 201: A first plurality of parcels may be obtained. For example, the first plurality of parcels is a plurality of parcels obtained, or otherwise received, at a warehouse, logistics center, or similar.

[0055] Action 202: Optionally, the parcels of the first plurality of parcels are scanned (e.g. optically read) to obtain first data relating to spatial dimensions of each one of the parcels in the first plurality of parcels.

[0056] Action 203: Optionally, second data relating to a source, such as a merchant, associated with the first plurality of parcels is also obtained.

[0057] Action 204: Based on the first data obtained in action 202 and / or the second data in action 203, the method may further comprise determining, or otherwise predicting, whether a parcel is either one of two types of parcels. Typically, a first type of parcel is an inflexible parcel, such as a box, and a second type of parcel is a flexible parcel, such as a bag. In other words, the method may include determining, or otherwise predicting, whether a parcel is an inflexible parcel (e.g., box) or a flexible parcel (e.g., bag) based on either scanned spatial dimensions of the parcel in question, or the source of the parcel, or a combination of both. This may be reduced into practice by utilizing a machine-learning (ML) algorithm for classifying each parcel of the plurality of parcels into either one of the said two types of parcels. For example, persons skilled in the art will appreciate that, at the time of filing this disclosure, it is conceivable to utilize either one or a combination of i) a K-nearest neighbours (KNN) classification algorithm, ii) Radius Neighbours classification algorithm and iii) a support vector machine algorithm. In addition, a logistic regression algorithm could advantageously be utilized for the parcel classification. Utilizing the above-mentioned ML algorithms for the purpose of classifying parcels is known per se and will therefore not be further detailed herein.

[0058] Action 205: A second plurality of parcels is obtained. For example, the second plurality of parcels may be collected from the first plurality of parcels that was obtained (received) in action 201. The second plurality of parcels is thus a subset of the larger first plurality of parcels. In some example implementations, action 202 is performed manually by an operator, or personnel, in the warehouse or logistics center. However, in advantageous implementations, action 202 is performed in a fully automated manner or in a semi-automated manner. In some implementations, e.g., robots are utilized to obtain (collect) the second plurality of parcels from the first plurality of parcels.

[0059] In advantageous implementations, the optional actions 202 through 204 have been performed prior to action 205 and action 205 then includes obtaining only parcels that are determined, or otherwise predicted, to be of one of the two types of parcels - typically flexible parcels. In other words, the second plurality of parcels would then be made up of only flexible parcels, such as bags.

[0060] Action 206: Spatial dimensions of each parcel of the obtained second plurality of parcels are determined, or otherwise sensed (action 202).

[0061] Action 207: Dependent on spatial dimensions of one or more separate dimensionally-constrained lockers of the electronic parcel compartment system 100, it is further determined in which dimensionally-constrained locker(s) a parcel can be placed. In some embodiments, this action may be performed manually, e.g., by an operator at a warehouse. For example, the operator may perform an ocular determination of the parcels to determine, or otherwise conclude, in which dimensionally-constrained locker(s) the parcels can be placed and, thus, fit. In alternative advantageous embodiments, this action is semi-automated or even fully or substantially fully automated. For example, automated processing systems may be provided in the warehouse or logistics center where said automated processing systems are configured to determine in which dimensionally-constrained locker(s) a parcel can be placed. The automated processing may e.g. involve optically reading (scanning) parcels and estimate weights, lengths, heights of flexible as well as inflexible parcels etc.

[0062] In some implementations, this action may involve utilizing ML algorithm. In other words, an ML algorithm may be utilized for assisting in the determination of which dimensionally-constrained locker(s) a parcel can be placed in. The machine learning algorithm may e.g. be one or a combination of: i) a KNN classification algorithm, ii) a weighted KNN classification algorithm, and iii) logistic regression algorithm.

[0063] Action 208 From action 207 it is made possible to know the smallest possible dimensionally-constrained locker of the electronic parcel compartment system in which a parcel in question can be placed, and thus fit. Based on this, the method involves assigning a value indicative of the smallest possible dimensionally-constrained locker of the electronic parcel compartment system in which it can be placed. In some example implementations, the values may be values 1 through 3.

[0064] Action 209: The parcels of the obtained second plurality of parcels are grouped into different groups of parcels based on the assigned values, see Table 1 below. In other words, different groups of parcels for each compartment size can be created or otherwise formed. In some embodiments, this action may be performed manually by an operator at the warehouse or logistics center. For example, the operator may put all the flexible parcels that would fit into a C8 sized compartment together in one group, and put all flexible parcels that would fit into a C12 sized compartment together in another group, and so on for all the different sizes. Table 1: Example of assigned values and corresponding groups.

[0065] In alternative advantageous embodiments, this action is semi -automated or even fully or substantially fully automated. For example, automated processing systems may be provided in the warehouse or logistics center to perform the grouping of parcels based on the assigned values. Similar to above, the automated processing system may be provided to create a group of all the flexible parcels that would fit into a C8 sized compartment, and create another group of all flexible parcels that would fit into a C12 sized compartment, and so on for all the different sizes.

[0066] Action 210: Based on the different groups that have been created or formed in action 209, a dataset is generated. This dataset is thus a dataset including data to the different groups of parcels based on the assigned values. This dataset may thus form knowledge of different flexible parcels. This dataset, and thus the assimilated knowledge about parcels and respective smallest possible dimensionally-constrained lockers of the electronic parcel compartment system in which they can be placed, can be used for subsequent training of a ML algorithm.

[0067] Turning now to Figs. 3-4, another aspect of the inventive concept of this disclosure will be described. Fig. 3 illustrates a flow chart of a illustrating a computer- implemented method 300 of training a ML algorithm. Fig. 4 illustrates an example user interface of a specific user interface of a sorting machine, which is tested and evaluated by the applicant of the present disclosure.

[0068] Reference is first made to Fig. 3, which shows a flowchart of method 300 of training a machine-learning algorithm. The method is computer implemented. That is to say, a computer system (e.g., an apparatus) comprising a memory and a processor may execute the actions, or method steps, of the method 300. In this example, the computer- implemented method 300 is implemented in an apparatus to model the matching of parcels with dimensionally-constrained lockers of an electronic parcel compartment system. Advantageously, but not necessarily, the apparatus may be a sorting machine for sorting parcels.

[0069] Action 301: The ML algorithm is initialized. Action 302: Once the ML algorithm has been initialized, a data set including data indicative of different groups of parcels based on assigned values is optionally obtained. The assigned values are indicative of a smallest possible dimensionally- constrained locker of the electronic parcel compartment system in which a parcel can be placed. Advantageously, but not necessarily, the data set has been generated by a method 200 as described in conjunction with Fig. 2.

[0070] Action 303: A plurality of parcels that are members of a first group of parcels and thus being capable of being placed in the same smallest possible dimensionally- constrained locker of the electronic parcel compartment system are provided, or otherwise collected.

[0071] Action 304: Spatial dimensions of the parcels of the plurality of parcels of the first group of parcels provided in action 303 are scanned, or otherwise sensed.

[0072] Next, actions 303-304 are repeated for n groups of parcels until all n groups of parcels have been scanned, or otherwise sensed, n is an integer number of 1 to an selectable maximum number, i.e., maximum number of groups of parcels. The exact value of n may be tested and evaluated for each specific use case in practice.

[0073] Action 305: The ML algorithm is updated. The ML algorithm may be updated after each repetition of actions 303-304 or, alternatively, when all n groups have been scanned, or sensed.

[0074] As will be appreciated, the computer-implemented method 300 can be repeated m number of times, m is an integer number of 1 to an selectable maximum number, i.e., maximum number of repetitions. The exact value of m may be tested and evaluated for each specific use case in practice.

[0075] In the following, a real world experiment performed by the applicant on an example implementation will be described in further detail in order to ease the understanding of certain aspects and embodiments of the invention as defined by the appended claims and described hitherto.

[0076] 1. Parcels of different types (such as boxes (inflexible) and bags (flexible)) arrive at a warehouse or logistics center.

[0077] 2. The parcels that have arrived at the warehouse or logistics center, are collected. The collected parcels are provided to, e.g. compartment size measurement equipment or similar. Parcels may be identified as either inflexible (e.g., boxes) or as flexible (e.g., bags). Once flexible parcels have been identified, or collected, these parcels are folded as these parcels would have been inserted or put in the locker of the electronic compartment system. The different parcels are assigned values depending on this and grouped into different groups of parcels for each compartment size. For example, all the flexible parcels that would fit into a C8 sized compartment are grouped together in one group, and all the flexible parcels that would fit into a C12 sized compartment are grouped together in another group, and so on for all the different sizes. Figs. 4A and B show an illustration and a corresponding photograph from the example experiment where parcels have been collected and grouped into three exemplary groups. Different groups of parcels are provided to a sorting machine. Fig. 5 schematically illustrates the user interface (UI) appearing at the sorting machine for an operator who interacts and operates the sorting machine. The sorting machine settings can be updated such that a machine learning setting is on and such that the correct group size is selected, e.g. C8 for the group that fits into the C8 compartment. The flexible parcels in the group in question are scanned, or otherwise sensed. While the machine learning setting is on, it will be appreciated that it is important that only parcels belonging to the same group of parcels are scanned. Items, or steps 5 and 6, are repeated for the next group of parcels, making sure to select the correct locker compartment size for this group at the UI. When all the parcel groups have been scanned, turn off the machine learning setting on the sorting machine by interacting and operating the corresponding virtual button of the UI (see Fig. 6). It will further be appreciated that it is important that nothing else is scanned by the sorting machine when the machine learning setting is turned on. Otherwise, one will run the risk of trying to put parcels into locker compartments that are of a wrong size. If wrong parcels are scanned, e.g. by mistake, or the wrong size is used for the machine learning on the sorting machine, the sorting machine is operable to override these to correct the sizes.

[0078] In more detail, the sorting machine may be capable of learning errors and take corrective measures based on the learned errors. For example, when a parcel is scanned the sorting machine has learned - by means of the ML - when a parcel of a certain size actually does fit into a certain box (or vice versa), the sorting machine can learn to override earlier conclusions made by the sorting machine to take corrective measures for future parcels.

[0079] Reference is first made to Fig. 7, which shows a flowchart of a method 700. The method 700 is computer-implemented. That is to say, a computer system (e.g., an apparatus) comprising a memory and a processor may execute the actions, or method steps, of the method 700. In this example, the computer-implemented method 300 is implemented in an apparatus for matching of a parcel with dimensionally-constrained lockers of an electronic parcel compartment system. Optionally, the apparatus may be embodied in a sorting machine for sorting parcels.

[0080] Action 701: A parcel is scanned, or otherwise sensed, to determine spatial dimensions of the parcel.

[0081] Action 702: The trained machine learning algorithm of described in conjunction with Fig. 3 is applied to the determined spatial dimensions of the parcel.

[0082] Action 703: Based on this, a value is produced or otherwise generated. This value is indicative of a probability that the scanned parcel can fit inside one or several of the dimensionally-constrained lockers of the electronic parcel compartment system.

[0083] This has the advantage that it enables a courier service to know in advance the probability that a scanned parcel will fit inside one or more lockers of an electronic parcel compartment system. By knowing this before deliveries is advantageous in that the deliveries can be planned more accurately and, in turn, this enables a facilitated delivery of parcels of different shapes and sizes to suitably sized lockers of an electronic parcel compartment system. The deliveries, and thus transports, can be planned more cleverly and fewer transports may require for delivering a certain amount of parcels. By way of extension, fewer and smarter transports may lead to a reduced environmental footprint by delivery services.

[0084] Example implementations of embodiments of the apparatus

[0085] Fig. 8 illustrates a data processing system, or apparatus 800, which is configured for performing or otherwise executing the methods according to the various aspects and embodiments discussed in this disclosure. As is schematically illustrated in Fig. 8, the apparatus 800 comprises hardware 810, 820, 830, 840. For example, the apparatus 800 may comprise one or more processors 810 and one or more memories 820. Also, a communications interface 830 may be provided in order to allow the apparatus 800 to communicate with other apparatuses and / or communication devices, e.g. via a network such as the Internet. To this end, the communications interface 830 may comprise a transmitter (Tx) and a receiver (Rx). Alternatively, the communications interface 830 may comprise a transceiver (Tx / Rx) combining both transmission and reception capabilities. The communications interface 830 may include a radio frequency (RF) interface allowing the data processing system 800 to communicate with other data processing systems and / or communication devices through a radio frequency band through the use of different radio frequency technologies such as 5G NR (New Radio), LTE (Long Term Evolution), WCDMA (Wideband Code Division Multiple Access), or any other cellular network standardized by the 3rd Generation Partnership Project (3GPP), or any other wireless technology such as Wi-Fi, Bluetooth®, etcetera. The data processing system 800 may optionally also comprise a user interface 840.

[0086] Example computer-readable mediums

[0087] Turning now to Fig. 9, another aspect will be briefly discussed. Fig. 9 shows an example of a computer-readable medium, in this example in the form of a data disc 1000. In one embodiment the data disc 900 is a magnetic data storage disc. The data disc 900 is configured to carry instructions 900 that can be loaded into a memory 820 of a data processing system, or apparatus 800. Upon execution of said instructions by a processor 810 of the apparatus 800, the apparatus 800 is caused to execute a method or procedure according to the embodiments disclosed in this disclosure. The data disc 900 is arranged to be connected to or within and read by a reading device (not shown), for loading the instructions into the processor. One such example of a reading device in combination with one (or several) data disc(s) 900 is a hard drive. It should be noted that the computer- readable medium can also be other mediums such as compact discs, digital video discs, flash memories or other memory technologies commonly used. In such an embodiment the data disc 900 is one type of a tangible computer-readable medium. The instructions may alternatively be downloaded to a computer data reading device, such as the apparatus 800 or another computer capable of reading computer coded data on a computer-readable medium, by comprising the instructions in a computer-readable signal (not shown) which is transmitted via a wireless (or wired) interface (for example via the Internet) to the computer data reading device for loading the instructions into a processor 810 of the data processing system 800. In such an embodiment, the computer-readable signal is one type of a non-tangible computer-readable medium.

[0088] Various aspects and embodiments described herein are advantageous in that they enable an improvement over the existing art. Some aspects and embodiments proposed herein allow for a facilitated delivery of parcels of different shapes and sizes to suitably sized lockers of an electronic parcel compartment system.

[0089] For example, the inventors have recognized that it is beneficial to know (in advance) a probability that a parcel will fit inside one or more lockers of an electronic parcel compartment system. By knowing this before deliveries, it is possible to plan deliveries more accurately and, in turn, this enables a facilitated delivery of parcels of different shapes and sizes to suitably sized lockers of an electronic parcel compartment system. The deliveries, and thus transports, can also be planned more cleverly and fewer transports may be required for delivering a certain amount of parcels. Fewer and smarter transports may also lead to a reduced environmental footprint by delivery services.

[0090] As will be appreciated by those skilled in the art, modifications and other variants of the described aspects and embodiments will come to mind to one skilled in the art having benefit of the teachings presented in the foregoing description and associated drawings. Therefore, it is to be understood that the embodiments are not limited to the specific example embodiments described in this disclosure and that modifications and other variants are intended to be included within the scope of this disclosure. For example, a person skilled in the art will appreciate that there exist various ways implementing the action, or method step 205 of obtaining the second plurality of parcels. In certain implementations, it is conceivable to merely determine, or otherwise predict, whether a parcel is either one of two types of parcels (i.e., an inflexible parcel such as a box or a flexible parcel such as a bag). In other words, it is not necessary, nor essential, to scan a first plurality of parcels to obtain first data for use in determining, or otherwise predicting, whether the parcel is either one of said two types of parcels. As one example implementation, it is conceivable to use a camera or other image processing device to classify pixels of a captured image for determining whether a parcel is either one of said two types of parcels. Furthermore, although specific terms may be employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation. Therefore, a person skilled in the art would recognize numerous variations to the described embodiments that would still fall within the scope of the appended claims. As used herein, the terms “comprise / comprises” or “include / includes” do not exclude the presence of other elements or steps. Furthermore, although individual features may be included in different claims, these may possibly advantageously be combined, and the inclusion of different claims does not imply that a combination of features is not feasible and / or advantageous. In addition, singular references do not exclude a plurality.

Claims

CLAIMS1. A method (200) of generating a dataset for a machine-learning algorithm, the method (200) comprising:(i) obtaining (205) a second plurality of parcels, which is a subset of larger first plurality of parcels;(ii) determining (206), or otherwise sensing, spatial dimensions of each one of the obtained second plurality of parcels;(iii) dependent on spatial dimensions of one or more separate dimensionally-constrained lockers of an electronic parcel compartment system, determining (207) in which dimensionally-constrained locker(s) a parcel can be placed;(iv) assigning (208) each parcel a value indicative of the smallest possible dimensionally-constrained locker of the electronic parcel compartment system in which it can be placed;(v) grouping (209) the parcels of the obtained second plurality of parcels into different groups of parcels based on the assigned values; and(vi) generating (210) a dataset including data relating to the different groups of parcels based on the assigned values.

2. The method (200) according to claim 1, further comprising, prior to (i) obtaining (205) the second plurality of parcels: scanning said first plurality of parcels to obtain (202) first data relating to spatial dimensions of each one of the parcels in the first plurality of parcels; and based on the first data determining (204), or otherwise predicting, whether a parcel is either one of two types of parcels; and wherein obtaining (205) the second plurality of parcels includes obtaining only parcels that are determined, or otherwise predicted, to be of one of the two types of parcels.

3. The method (200) according to claim 2, further comprising, prior to (i) obtaining (205) the second plurality of parcels: obtaining (203) second data relating to a source, such as a merchant, associated with the first plurality of parcels; and based on a combination of the first and second data determining (204), or otherwise predicting, whether a parcel is either one of two types of parcels; and wherein obtaining (205) the second plurality of parcels includes obtaining only parcels that are determined, or otherwise predicted, to be of one of the two types of parcels.

4. The method (200) according to claim 2 or 3, wherein: a first type of parcel is an inflexible parcel, such as a box; a second type of parcel is a flexible parcel, such as a bag; and wherein obtaining (205) the second plurality of parcels includes obtaining only parcels that are determined, or otherwise predicted, to be of the second type.

5. The method (200) according to any one of the claims 2-4, wherein determining (204), or otherwise predicting, whether a parcel is either one of two types of parcels includes: utilizing a machine-learning algorithm for classifying each parcel of the plurality of parcels into either one of the two types of parcels.

6. The method (200) according to claim 5, comprising utilizing one or a combination of:- a K-nearest neighbours, KNN, classification algorithm;- a Radius Neighbours classification algorithm; and- a support vector machine algorithm.

7. The method (200) according to claim 6, further comprising utilizing:- a logistics regression algorithm.

8. The method (200) according to any of the claims 1-7, wherein (iii) determining (207) in which dimensionally-constrained locker(s) a parcel can be placed comprises: utilizing a machine-learning algorithm, wherein the machine learning algorithm utilizes one or a combination of:- a KNN classification algorithm;- a weighted KNN classification algorithm; and- a logistics regression algorithm.

9. A computer-implemented method (300) of training a machine-learning algorithm implemented in an apparatus to model the matching of parcels with dimensionally-constrained lockers of an electronic parcel compartment system, the method (300) comprising:(a) initialising (301) a machine learning algorithm;(b) obtaining (302) a data set including data indicative of different groups of parcels based on assigned values, wherein the assigned values are indicative of a smallest possible dimensionally-constrained locker of the electronic parcel compartment system in which a parcel can be placed;(c) providing (303) a plurality of parcels being members of a first group of parcels and thus being capable of being placed in the same smallest possible dimensionally-constrained locker of the electronic parcel compartment system;(d) scanning (304) spatial dimensions of the parcels of the plurality of parcels of the first group of parcels;(e) repeating (c)-(d) for n groups of parcels until all n groups of parcels have been scanned;(f) updating (305) the machine learning algorithm accordingly.

10. The computer-implemented method (300) of claim 9, wherein the obtained (302) data set has been generated by a method according to any one of the claims 1-8.

11. The computer-implemented method (300) of claim 9 or 10, wherein the computer-implemented method (300) is repeated m number of times.

12. A computer-implemented method (700) implemented in an apparatus for matching of a parcel with dimensionally-constrained lockers of an electronic parcel compartment system, the computer-implemented method (700) comprising: scanning (701) a parcel to determine spatial dimensions of the parcel; applying (702) the trained machine learning algorithm of any one of the claims 9-11 to the determined spatial dimensions of the parcel; and based on this producing (703) a value indicative of a probability that the scanned parcel can fit inside one or several of the dimensionally-constrained lockers of the electronic parcel compartment system.

13. An apparatus (800) comprising means (810, 820) adapted to carry out any one of the methods according to any one of the claims 1-12.

14. The apparatus (800) according to claim 13, wherein the apparatus is a sorting machine for sorting parcels.

15. A computer program comprising instructions (910) which, when executed on at least one processor (810), cause the at least one processor (810) to carry out any one of the methods according to any one of the claims 1-12.

16. A carrier comprising the computer program according to claim 15, wherein the carrier is any one of an electronic signal, an optical signal, a radio signal, or a computer-readable medium (900).

Images