Training for technical specialists

Abstract

Training software for human operators (1A), for example for a pilot student (5), to teach him or her to accomplish a training activity, for example a piloting training activity; the training software for human operators (1A) is storable in and executable by electronic processing resources (2) and designed to cause, when executed, said electronic processing resources (2) to become configured to cause the display (block 14), by means of electronic display or projection resources (3), of a rendered virtual working environment (7) corresponding to a real working environment in which the student (5) will have to operate; and to receive (block 14) from an electronic controller (4), operable by a student (5), a number of commands imparted by the student (5) and indicative of one or different actions to be performed by a virtual avatar (6) of the student (5) in the rendered virtual working environment (7). Furthermore, the training software for human operators (1A) is designed to cause, when executed, said electronic processing resources (2) to become configured to cause the display (block 14), by means of the electronic display or projection resources (3) and in response to, and on the basis of, the commands received from the electronic controller (4), of the execution of the actions associated with these commands and performed by the virtual avatar (6) of the student (5) in the rendered virtual working environment (7). Furthermore, the training software for human operators (1A) is designed to cause, when executed, the electronic processing resources (2) to become configured to implement a simulation model (8) configured to determine an action, and / or a sequence of actions, to be performed in order to accomplish a predefined training activity, in particular a piloting activity, based on a plurality (80) of actions and / or sequences of actions and based on evaluation scores (81) associated with them; wherein, the evaluation scores (81) are indicative of the contribution of such actions, or sequences of actions, (80) to the accomplishment of the predefined training activity in the rendered virtual working environment (7). The training software for human operators (1A) is further designed to cause, when executed, the electronic processing resources (2) to become configured to determine a number of actions to be performed, in order to accomplish a training activity to be learned, based on one or more actions performed by the virtual avatar (6) of the student (5) in the rendered virtual working environment (7), and on the basis of the simulation model (8) implemented. Furthermore, the training software for human operators (1A) is designed to cause, when executed, the electronic processing resources (22) to become configured to cause the display (block 14), by means of the electronic display or projection resources (3), of a teaching representation configured to guide the student (5) in learning the determined actions to be performed in order to accomplish the training activity to be learned in the rendered virtual working environment (7).

Description

[0001] TRAINING FOR TECHNICAL SPECIALISTS

[0002] Cross-Reference To Related Applications

[0003] This patent application claims priority from Italian patent application no . 102024000002253 filed on February 2 , 2024 , the entire disclosure of which is incorporated herein by reference .

[0004] Technical Field of the Invention

[0005] The present invention concerns in general the instruction of students , more speci fically the training of pilots . In particular, the present invention concerns a computer system to allow a pilot student to autonomously learn one or more activities required by a training course .

[0006] State of the Art

[0007] As is known, the training of a specialist technician, in particular a pilot , normally takes place under the guidance of quali fied instructors who teach and observe the actions of the student . In detail , the instructors evaluate and correct the actions carried out by the student , or pilot , according to a predetermined ability obj ective or predefined instruction level .

[0008] Possibly, the instructors evaluate and correct the actions carried out by the student also in consideration of the abilities acquired by the student and expediently also those not yet acquired .

[0009] In further detail , the instructors guide the student in his / her training course , furthermore taking account of the learning history and progress of the student .

[0010] The use of simulation systems for simulating the piloting activity in order to mitigate cost and safety problems in the training of an aircraft pilot is known .

[0011] It is known that the use of simulation systems has considerable advantages with respect to the piloting of aircraft in a real environment ; in particular, said systems allow safe training of the pilot and reduction of the training costs .

[0012] For the above reasons , the use of simulation systems for simulating piloting activities plays a very important role in pilot training . Simulation systems in virtual reality are also known, via which a pilot can experience immersive flying in complete safety .

[0013] With the development of flight simulation technology and virtual reality technology, it is known that the existing flight simulators can achieve an extremely high simulation fidelity .

[0014] It is also known that CN115222300A describes an aviation simulator retraining time length distribution system and method based on an intelligent evaluation algorithm . According to the method, a virtual instructor i s provided and the technical shortcomings of the pilot student are determined according to the deviation between the performance of the pilot in a preevaluation operation and that of the virtual instructor . According to the method described, a regular polygon evaluation method is adopted, according to which pilots who have similar problems can be quickly identi fied based on the vector direction, and a quick grouping is performed . According to this method, the instructor operating curve is adopted as the evaluation index . The method described also allows the di f ferent driver habits of a pilot to be taken into consideration .

[0015] In addition, US 2022 / 139252 Al describes a method for evaluating the performance of a student during a training session in a training device . The method described includes receiving data, comparing the performance of the student with a model and assigning a score to the student . In one of the embodiments , the training device is a flight simulator configured to teach the student to pilot an aircraft . The flight simulator shows the output to the student and receives an input from the student . Furthermore , US 2023 / 297888 Al describes an automatic learning system for training students which comprises a first adaptive training system provided with an arti ficial intelligence module to adapt the individuali zed training to a first group of students and develop a learning model based on a set of learning performance metrics . The automatic learning system for student training described further comprises a second adaptive training system that provides individuali zed training to a second group of students and has a data property extraction module for extracting statistical properties from the learning performance metrics of the second group of students . In detail , the described automatic learning system for training students further comprises a data simulation module that generates simulated performance metrics using the extracted statistical properties to create a second learning model . Furthermore , the described automatic learning system for training students comprises a federated computing device which receives the weights of the learning models and generates or refines a federated model based on these weights .

[0016] Object and Summary of the Invention

[0017] The Applicant has observed that the known solutions are subj ect to improvement .

[0018] In particular, the Applicant has observed that the teaching, evaluation and correction processes implemented by a human instructor are inevitably guided by a subj ective component .

[0019] The obj ect of the present invention is therefore to make available a training software for human operators , in particular aircraft pilots , which improves , at least partly, the solutions of the known art ; in particular, completely replacing the subj ective evaluation of the instructor with an obj ective evaluation and providing a student with indications that are non-sub j ective and, at the same time , suited to the ability level demonstrated by the student in question .

[0020] The obj ect of the present invention is also to make available a training software for human operators that allows the student or human operator to be instructed, in a detailed in-depth manner, based on what the latter has done previously; namely, based on the actions and operations which he / she performed previously .

[0021] According to the present invention a training software for human operators as claimed in the attached claims is made available .

[0022] According to the present invention, a training software for human operators , for example a pilot student , is made available to teach him / her to accomplish a training activity, for example a piloting training activity; the training software for human operators can be stored in and executed by electronic processing resources and designed to cause , when executed, said electronic processing resources to become configured to cause the display, by means of electronic display or proj ection resources , of a rendered virtual working environment corresponding to a real working environment in which the student will have to operate ; and to receive from an electronic controller, operable by a student , a number of commands imparted by the student and indicative of one or di f ferent actions to be accomplished by a virtual avatar of the student in the rendered virtual working environment . Furthermore , the software for training human operators is designed to cause , when executed, the electronic processing resources to become configured to cause the displaying, by means of the electronic display or proj ection resources and in response to , and on the basis of , the commands received from the electronic controller, of the execution of the actions associated with said commands and carried out by the virtual avatar of the student in the rendered virtual working environment . Furthermore , the software for training human operators is designed to cause , when executed, the electronic processing resources to become configured to implement a simulation model configured to determine an action, and / or a sequence of actions , to be performed in order to accomplish a predefined training activity, in particular a piloting activity, based on a plural ity of actions and / or sequences of actions and based on evaluation scores associated with them; in which the evaluation scores are indicative of the contribution of such actions , or sequences of actions , to the accomplishment of the predefined training activity in the rendered virtual working environment . The training software for human operators is further designed to cause , when executed, the electronic processing resources to become configured to determine a number of actions to be performed, in order to accomplish a training activity to be learned, based on one or more actions performed by the virtual avatar of the student in the rendered virtual working environment , and based on the simulation model implemented . Furthermore , the training software for human operators is designed to cause , when executed, the electronic processing resources to become configured to cause the displaying, by means of the electronic display or proj ection resources , of a teaching representation configured to guide the student in learning the determined actions to be performed in order to accompl ish the training activity to be learned in the rendered virtual working environment .

[0023] Expediently, the simulation model is configured to receive information in input indicative of the actions performed by the virtual avatar of the student , to predict an evaluation score for a number of actions , which the virtual avatar of the student could performed, identi fied based on the performed actions ; and to determine and supply in output an action or sequence of actions to be performed based on a number of predicted evaluation scores .

[0024] Furthermore , optionally, the simulation model is configured to determine an action, and / or a sequence of actions , to be performed in order to accomplish a predefined training activity also based on the predefined training activity .

[0025] Expediently, the simulation model , for each of several predefined training activities , comprises a simulation sub-model associated with it . Furthermore , each of the simulation submodels is configured to determine an action, and / or a sequence of actions to be performed, to accomplish the predefined training activity associated with it , starting from the performed actions received in input .

[0026] The simulation model comprises at least one simulation neural network trained on a plurality of sample actions , and / or sequences of sample actions and evaluation scores associated with them, to determine an action, and / or a sequence of actions to be performed based on the actions performed by the virtual avatar of the student received in input .

[0027] Preferably, the simulation neural network is a recurrent neural network (RNN) trained to capture one or several relationships in a sequence of actions , comprising the actions accomplished and one or more actions to be performed in sequence , in order to predict the evaluation score of said sequence of actions .

[0028] Furthermore , according to an aspect of the present invention, the training software for human operators is designed to cause , when executed, the electronic processing resources to become further configured to implement or receive an evaluation model configured to ass ign an evaluation score to an action, and / or a sequence of actions , to be performed in order to accomplish a predefined training activity based on a plurality of actions and / or sequences of actions and based on evaluation scores associated with them .

[0029] Furthermore , according to said aspect of the present invention, the electronic processing resources become further configured to determine and assign an evaluation score to one or more of the actions performed by the virtual avatar of the student , in order to accomplish at least partly the predefined training activity, based on the evaluation model . The electronic processing resources become further configured to cause the displaying, by means of the electronic display or proj ection resources , of a teaching representation, indicative of a learning progress stage of the student in relation to the predefined training activity, based on a number of evaluation scores assigned to one or more actions performed in the rendered virtual working environment in order to accomplish the predefined training activity .

[0030] Optionally, the training software for human operators is further designed to cause , when executed, the electronic processing resources to become configured to cause the display, by means of the electronic display or proj ection resources , of a graphic representation indicative of the need to , or requiring the student to , repeat execution of the predefined training activity, or at least part of the predefined training activity, or continue with the execution of a di f ferent action, or training activity, with respect to the performed action, or the training activity, predef ined based on the evaluation score assigned to the performed action .

[0031] According to an aspect of the present invention, the training software for human operators is furthermore designed to cause , when executed, the electronic processing resources to become configured to determine whether an action performed by the virtual avatar of the student in the rendered virtual working environment is , or is not , a killer action or an action harmful for the predefined training activity .

[0032] According to said aspect of the present invention, the training software for human operators is further designed to cause , when executed, the electronic processing resources to become configured to cause the displaying, by means of the electronic display or proj ection resources , of a graphic representation indicative of the need to , or requiring the student to , repeat execution of the predefined training activity, at least partly accomplished, i f it is determined that said action is a killer action .

[0033] Brief Description of the Drawings

[0034] Figure 1 shows a block diagram of a training system for human operators according to the present invention .

[0035] Figure 2 shows a flow chart representing a method that can be implemented by means of a training software for human operators according to an embodiment of the present invention .

[0036] Figure 3 shows a flow chart representing a method that can be implemented by means of the training software for human operators of Figure 2 .

[0037] Disclosure of Preferred Embodiments of the Invention

[0038] The present invention will now be described in detail with reference to the attached figures to enable a person skilled in the art to produce it and use it . Various modi fications to the embodiments described will be immediately evident to persons skilled in the art and the general principles described can be applied to other embodiments and applications without departing from the protective scope of the present invention, as defined in the attached claims . Therefore , the present invention shall not be considered limited to the embodiments described and illustrated, but shall be given the widest protective scope in accordance with the characteristics described and claimed .

[0039] Where not defined otherwise , all the technical and scienti fic terms used here have the same meaning as the one commonly used by persons having ordinary experience in the sector pertaining to the present invention . In the case of conflict , the present description, including the definitions provided, will prevail . Furthermore , the examples are provided for purely illustrative purposes and as such shall not be considered limiting .

[0040] In particular, the block diagrams included in the attached figures and described below shall not be understood as a representation of the structural characteristics , namely construction limitations , but shall be interpreted as a representation of functional characteristics , namely intrinsic properties of the devices defined by the ef fects obtained, namely functional limitations which can be implemented in di f ferent ways , therefore in such a way as to protect the functionalities (possibility of functioning) thereof .

[0041] In order to facilitate understanding of the embodiments described here , reference will be made to some speci fic embodiments and a speci fic language will be used to describe the same .

[0042] The terminology used in the present document has the purpose of describing only particular embodiments , and is not intended to limit the scope of the present invention .

[0043] Figure 1 shows a block diagram of a training system for human operators according to the present invention .

[0044] In particular, the training system for human operators , for example a pilot student 5 , is designed to teach said operators to accomplish a training activity, for example a piloting training activity; expediently, a series of training activities of a predefined training course .

[0045] It should be further noted that from here onwards the term training activity, or task, conveniently refers to a programme , or a lesson, aimed at developing knowledge or one or more skills or abilities in a student ; in particular, reference is made to the final result to be obtained at the end of a lesson or programme . By way of non-limiting example , a training activity could be an activation of the aircraft undercarriage system; therefore , at the end of the lesson, the student 5 must be able to activate the undercarriage system .

[0046] Furthermore , from here onwards , the term action conveniently refers to a speci fic particular action to be performed expediently, but without limitation, within a training activity . By way of non-limiting example , an action could be the interaction ( for example , by exerting a pressure ) with a speci fic push-button on the console of a reference station of the student 5 , which contributes , i f combined with other buttons , to activation of the aircraft undercarriage system .

[0047] In particular, the training system for human operators comprises electronic display or proj ection resources 3 designed to render, or to display (block 14 ) , a rendered virtual working environment 7 corresponding to a real working environment in which a student 5 will have to operate . Furthermore , the training system for human operators comprises an electronic controller 4 operable by the student 5 to impart a number of commands indicative of one or several actions to be performed by a virtual avatar 6 of the student 5 in the rendered virtual working environment 7 .

[0048] Furthermore , said training system for human operators further comprises electronic processing resources 2 storing, and designed to execute , a software (namely a computer product ) for training human operators 1A.

[0049] In detail , the training software for human operators 1A is storable in and executable by the electronic processing resources 2 and is designed to cause , when executed, said electronic processing resources 2 to become configured to allow the student 5 to learn and / or train to accomplish a predefined training activity . In particular, the electronic processing resources 2 are configured to allow the student 5 to learn or train in, in guided autonomy, a number of predefined training activities of a training course .

[0050] The electronic processing resources 2 allow the student 5 to try out and learn a multitude of actions , in order to accomplish a training activity, without placing limits on the actions which the student 5 can perform . In further detail , the aim is for the sel f-training of the student 5 to proceed via a number of training obj ectives , or activities , to be achieved one at a time .

[0051] With regard to the learning or training of the student 5 , what counts are the operations that have to be implemented to achieve said functionalities and not the hardware and software architecture with which said operations are implemented; in fact the latter could be implemented by means of a concentrated architecture , namely by a single electronic device (by way of example , a single server ) , or by means of cooperative distributed architecture , namely distributed among di f ferent electronic devices in communication and cooperating with one another according to a proprietary logical architecture which the producer of the training software for human operators 1A will decide to adopt .

[0052] The electronic processing resources 2 are designed to cause the display (block 14 ) , by means o f the electronic display or proj ection resources 3, of a rendered virtual working environment 7 corresponding to a real working environment in which the student 5 will have to operate . In particular, the electronic display or proj ection resources 3 are configured to immerse the student 5 , expediently with a plurality of senses ( such as sight , hearing and touch) , in the rendered virtual working environment 7 .

[0053] By way of example , said working environment , whether real or virtual , comprises a set of instruments and sensors speci fic for the development of a required activity ( or task) , a reference station for the student 5 where he / she can carry out some functionalities , expediently but without limitation by means of manual operations , and a series of components similar in type to the nature of the activity to be carried out ; preferably, said components are not necessary for the execution of said activity but are available for testing the student 5 vis-a-vis the execution of actions that can compromise her / his training .

[0054] In further detai l , the electronic processing resources 2 are designed to communicate and / or cooperate (block 14 of Figure 2 ) with the electronic display or proj ection resources 3 in order to cause the displaying (block 14 ) of said virtual working environment 7 .

[0055] According to an embodiment of the present invention, the electronic display or proj ection resources 3 comprise a virtual reality viewer that can be worn by the student 5 .

[0056] By way of non-l imiting example , the virtual reality viewer comprises one or more screens , for example LCD ( Liquid Crystal Display) or OLED ( Organic Light-Emitting Diode ) screens , arranged to have images displayed (block 14 ) and expediently to provide a three-dimensional perception of what is intended to be represented, one or more tracking sensors for monitoring the body movements ( optionally the head and preferably also the hands ) of the student 5 , and audio devices or headsets to allow the student 5 to perceive or hear one or more sound contents ; furthermore , the virtual reality viewer optionally implements the foveated rendering technology to adj ust the graphic resolution based on the direction of the gaze of the student 5 .

[0057] According to a di f ferent embodiment of the present invention, the electronic display or proj ection resources 3 comprise one or more from among a screen of a computer, a screen of a smartphone , of a tablet or phablet , or a proj ector to proj ect one or several images on one or more walls so that said images can be viewed by the student 5 .

[0058] Furthermore , the electronic processing resources 2 are designed to receive (block 14 of Figure 2 ) from the electronic controller 4 , operable by the student 5 , a number of commands imparted by the student 5 and indicative of one or several actions to be performed by a virtual avatar 6 of the student 5 (namely, a digital representation in first or third person of a student 5 ) in the rendered virtual working environment 7 . According to an aspect of the present invention, the virtual reality viewer comprises the electronic controller 4 configured to receive one or more quantities measured by a number of tracking sensors of the viewer and / or by a number of tactile or pressure sensors (by way of example , gloves or other virtual reality equipment ) and to determine one or more commands imparted by the student 5 based on said quantities received .

[0059] According to a di f ferent aspect of the present invention, the electronic controller 4 is a controller external to the virtual reality viewer, by way of example with push-buttons and / or analogue levers with which the student 5 can interact to impart one or more commands .

[0060] The electronic processing resources 2 are further designed to cause the display (block 14 ) , by means of the electronic display or proj ection resources 3 and in response to , and on the basis of , the commands received from the electronic controller 4 , of the execution ( or performance ) of the actions associated with said commands and performed by the virtual avatar 6 of the student 5 in the rendered virtual working environment 7 .

[0061] In detail , the electronic display or proj ection resources are designed to graphically simulate the execution o f a training activity by the virtual avatar 6 of the student 5 in the virtual working environment 7 . In further detail , according to an aspect of the present invention, the electronic processing resources 2 are configured to determine a number of actions corresponding to one or more commands imparted by the student 5 by means of the electronic controller 4 based on a predefined mapping between commands and actions to correlate commands which a student 5 could impart to a number of actions to be performed in the virtual environment .

[0062] Figure 2 shows a flow chart representing a method implementable by means of a training software for human operators according to an embodiment of the present invention .

[0063] The electronic processing resources 2 are further configured to implement a simulation model 8 configured to determine an action, and / or a sequence of actions , to be performed in order to accomplish a predefined training activity, in particular a piloting activity, based on a plurality of actions and / or sequences of actions 80 and based on evaluation scores 81 associated with them; in which the evaluation scores 81 are indicative of the contribution of said actions 80 , or sequences of actions , to the accomplishment of the predefined training activity in the rendered virtual working environment 7 .

[0064] According to an aspect of the present invention, an action is a word or a token, namely a textual content unit ; therefore , the sequences of actions are sequences of words .

[0065] In detail , the simulation model 8 is configured to receive information in input indicative of the performed or past actions ( for example , a textual content containing the list of the performed actions ) performed by the virtual avatar 6 of the student 5 .

[0066] Furthermore , said simulation model 8 is configured to predict an evaluation score , by way of example a numerical value comprised in a predefined range (by way of example , a range defined to comprise the values from - 1 to 1 ) , for a number of actions which the virtual avatar 6 of the student 5 could perform, identi fied based on the performed actions . The simulation model 8 is designed to determine and output an action or sequence of actions to be performed, or future actions , based on a number of , or all the , predicted evaluation scores .

[0067] In particular, the simulation model 8 is configured to determine , according to a sequence of performed or past actions received in input , the sequence of actions to be performed which achieves the maximum value in terms of evaluation scores with respect to the actions that can be performed starting from, or in sequence with, the performed actions .

[0068] In further detail , the simulation model 8 is further configured, or trained, to determine what actions can be performed based on the actions received in input and optionally also based on the predefined training activity .

[0069] In detail , the simulation model 8 is designed to receive in input an action or a sequence of actions and to determine an action or sequence of actions which the virtual avatar 6 of the student 5 could perform, based on the predefined training activity, which optimi zes or maximi zes the associated evaluation score ; in particular, said determined action or sequence of actions is associated with a higher evaluation score than any other action or sequence of actions which the virtual avatar 6 of the student 5 could perform . In further detail , the simulation model 8 is designed to determine or predict the evaluation score associated with an action or sequence of actions which the virtual avatar 6 of the student 5 could perform and to identi fy the action or the sequence of actions that maximi zes the evaluation score .

[0070] In particular, the simulation model 8 is configured to determine , or select , the subsequent action to be performed starting from the performed actions whose evaluation score is determined to be higher, considering the entire sequence of actions ( in which said sequence comprises the performed actions and one or more actions to be performed) , than the other actions which the simulation model 8 is able to simulate .

[0071] Furthermore , the simulation model 8 is designed to provide in output the action or sequence of actions determined and possibly the evaluation score associated with it .

[0072] According to a preferred embodiment of the present invention, the simulation model 8 comprises at least one neural simulation network trained, on a plurality of sample actions , and / or sequences of sample actions 80 , and evaluation scores 81 associated therewith, to determine an action, and / or a sequence of actions to be performed based on an input , or based on the actions performed by the virtual avatar 6 of the student 5 . In detail , the electronic processing resources are configured to train (block 82 ) the simulation model 8 , on a plurality of sample actions , and / or sequences of sample actions 80 and evaluation scores 81 associated therewith, to predict the sequence of actions to be performed optimi zing the evaluation score associated therewith .

[0073] In particular, the neural simulation network is trained on a set of training data 11 comprising (block 12 ) sequences of actions 80 and associated evaluation scores 81 .

[0074] According to an aspect of the present invention, the neural networks of the simulation model 8 are trained (block 82 ) by means of a supervised learning technique in which the sample data comprise di f ferent sequences of sample actions 80 of one or several predefined training activities and optionally further predefined parameters ( for example , an obj ective training level of the student 5 ) (block 13 ) , and the labels , associated with the sample data, comprise di f ferent evaluation scores 81 and optionally indexes of harmfulness (block 13 ) of the action (by way of example , a value indicative of the fact that the action is a killer action or not ) .

[0075] Optionally, the sample data comprise di f ferent sequences of sample actions 80 of one or several aircraft piloting training activities .

[0076] Optionally, the labels further comprise optimal sequences of actions , namely with an associated evaluation score higher than an evaluation threshold, which it is desirable to provide in output .

[0077] According to an aspect of said embodiment , at least one (possibly all ) neural simulation network i s a recurrent neural network (RNN) trained to capture one or several relationships in a sequence of actions , comprising in sequence the performed actions and one or more actions to be performed, in order to predict the evaluation score of said sequence of actions .

[0078] In particular, the neural simulation network is trained to determine the evaluation score associated with an action or a sequence of actions and to output the latter i f said evaluation score meets a predefined condition, in particular i f the evaluation score is not inferior to a di f ferent evaluation score associated with a di f ferent action or sequence of actions which the virtual avatar 6 of the student 5 could perform .

[0079] In particular, the simulation model 8 is a neural model comprising a plurality of neural networks trained on the basis of sample actions and / or sequences of sample actions 80 and based on the evaluation scores 81 assigned to said actions / sequences of actions 80 and indicative of the contribution of the latter to the success ful accomplishment of the training activity . In particular, the trained neural networks of the simulation model 8 are recurrent neural networks (RNN) ; possibly, they are LSTM ( long short-term memory) neural networks .

[0080] Unlike the traditional neural networks , the LSTM neural networks are provided with a long-term memory that allows them to capture long-range dependencies in sequences , making them particularly suitable for tas ks involving sequential data such as natural language recognition, automatic translation and many other applications .

[0081] In detail , the simulation model 8 is configured to determine an action, and / or a sequence of actions , to be performed in order to accomplish a predefined training activity, also based on the predefined training activity, or the type of training activity .

[0082] In particular, the simulation model 8 , for each of various predefined training activities , comprises a simulation sub-model 8 associated with it ; furthermore , each of the simulation submodels is configured to determine an action, and / or a sequence of actions , to be performed in order to accomplish the predefined training activity associated with it . In particular, each of said simulation sub-models comprises , or consists of , a speci fic simulation arti ficial neural network for a training activity, trained on actions and sequences of actions 80 speci fic for the training activity, to determine and to output a sequence of actions for said training activity .

[0083] In detail , each of the various sub-models is associated with, and i f necessary trained on, a di f ferent training activity .

[0084] Expediently, each sub-model is trained in a di f ferent set of data 11 , or on a di f ferent set of actions 80 and sequences of actions 81 (block 12 ) , speci fic for the training activity, which the virtual avatar 6 could perform . Furthermore , according to an embodiment of the present invention, the simulation model 8 comprises an orchestrator module configured to determine or classi fy the training activity, namely select the activity from among di fferent predefined training activities , and the sub-model associated with said training activity .

[0085] Furthermore , the orchestrator module is designed to command a simulation of the training activity to the associated sub-model and to receive the action or sequence of actions to be performed by said sub-model .

[0086] In particular, the electronic processing resources 2 are further configured to implement or receive an evaluation model 9 configured to evaluate or assign an evaluation score to an action, and / or a sequence of actions , to be performed in order to accomplish a predefined training activity based on a plurality of actions and / or sequences of actions 80 and based on evaluation scores 81 associated with them .

[0087] In particular, the evaluation model 9 is a mapping (by way of non-limiting example , a table or a list of pairs of values ) , or a function or method for performing a mapping, between actions or sequences of actions and evaluation scores .

[0088] According to an aspect of the present invention, the simulation model 8 comprises the evaluation model 9 ; in particular, the evaluation model 9 is a sub-model of the simulation model 8 .

[0089] Optionally, the evaluation model 9 comprises a neural evaluation network trained to determine and assign an evaluation score to an action or sequence of actions in input .

[0090] Furthermore , the electronic processing resources 2 are designed to determine and assign an evaluation score to one or more of the actions performed in the rendered virtual working environment 7 by the virtual avatar 6 of the student 5 , in order to accomplish at least partial ly the predefined training activity, based on the evaluation model 9 implemented .

[0091] In detail , the electronic processing resources 2 are designed to compute or determine the evaluation score of an action providing in input to the evaluation model 9 the action to be evaluated and pos sibly, but preferably, a number of actions , in sequence , performed in one or more times preceding the time at which the action to be evaluated was performed .

[0092] According to an optional aspect of the present invention, the electronic processing resources 2 are designed to generate and / or to output the set of data 11 for training the simulation model 8 via use of the evaluation model 9. In particular, the electronic processing resources 2 are designed to determine , via use of the evaluation model 9, the evaluation score for each sequence of actions in input and to store said evaluation score so that it is associated with the corresponding sequence of actions .

[0093] The electronic processing resources 2 are designed to determine a number of actions to be performed, in order to accomplish a training activity to be learned, based on one or more actions performed by the virtual avatar 6 of the student 5 in the rendered virtual working environment 7 , and based on the simulation model 8 implemented . In detail , the electronic processing resources 2 are configured to provide in input the performed actions to the simulation model 8 and are configured to receive an action or a sequence of actions to be performed, or future actions , by the implemented simulation model 8 . In further detail , the electronic processing resources 2 are configured to determine the actions to be performed for which the associated evaluation score meets a predefined condition; preferably, i f the evaluation score is higher than a di f ferent sequence of actions that could be performed . Furthermore , the electronic processing resources 2 are configured to determine and assign an evaluation score to one or more of the actions performed by the virtual avatar 6 of the student 5 , in order to accomplish at least partially the predefined training activity, based on the evaluation model 9.

[0094] Figure 3 shows a flow chart representing a method that can be implemented by means of the training software for human operators 1A shown in Figure 2 .

[0095] Furthermore , the electronic processing resources 2 are designed to cause the di splay (block 14 ) , by means of the electronic display or proj ection resources 3, of a teaching representation, indicative of a learning progress stage of the student 5 in relation to the training activity predefined and / or configured to guide the student 5 to perform ( or carry out ) at least part of the predefined training activity, or a di f ferent training activity, in the rendered virtual working environment 7 . In detail , the electronic processing resources 2 are configured to determine or compute the learning progress stage of the student 5, with respect to the activity to be accomplished in the virtual working environment 7 and / or with respect to a training course , based on a number of evaluation scores assigned to one or more actions , or to a sequence of actions , performed in the virtual working environment 7 in order to accomplish said predefined activity .

[0096] In further detail , the learning progress stage of the student 5 is indicative of the level of ability or learning of the student 5 in carrying out one or more predefined activities ; optionally, it is also indicative of the progress ( or regression) , over time , of the student 5 in learning or training of the predefined activities to be learned .

[0097] In particular, the electronic processing resources 2 are configured to cause the displaying (block 14 ) , by means of the electronic display or proj ection resources 3, of a teaching representation, indicative of the learning progress stage of the student 5 in relation to the predefined training activity, based on a number of evaluation scores assigned to one or more actions performed in the rendered virtual working environment 7 in order to accomplish the predefined training activity; in particular, based on an evaluation score assigned to the sequence of actions performed by the virtual avatar 6 of the student 5 .

[0098] The electronic processing resources 2 are designed to cause the displaying (block 14 ) , via the electronic display or proj ection resources 3, of a teaching representation configured to guide the student 5 in learning the actions determined and to be performed in order to accomplish the training activity to be learned in the rendered virtual working environment 7 . In particular, said teaching representation is a sequence of representations of steps , or movements , to be performed (by way of non-limiting example , represented by multi-dimensional images ) to perform the actions to be performed determined by means of the simulation model 8 .

[0099] In particular, the electronic processing resources 2 are configured to cause the display (block 14 ) , by means of the electronic display or proj ection resources 3, of a graphic representation indicative of the need to , or requiring the student 5 to , repeat execution of the predefined training activity, or at least part of the predefined training activity, or continue with the execution of a di f ferent action, or training activity, with respect to the performed action, or the training activity, predef ined based on the evaluation score assigned to the performed action .

[0100] According to the preferred embodiment , the electronic processing resources 2 are configured to cause the display (block 14 ) , by means of the electronic display or proj ection resources 3 and based on the evaluation scores assigned to the actions performed in the rendered virtual working environment 7 , of the teaching representation to guide the student 5 in learning at least part of a training activity (by way of example , the predefined training activity) to be accomplished by performing a number of predetermined actions in the rendered virtual working environment 7 ; and / or a representation indicative of the success , or failure , of at least part of the learning of the predefined training activity at least partially accomplished .

[0101] According to an aspect of the present invention, the electronic processing resources 2 are configured to cause the display (block 14 ) , via the electronic display or proj ection resources 3, of a graphic representation indicative of the need to , or requiring the student 5 to , repeat execution of the predefined training activity, or repeat the execution of at least part of the predefined training activity, or continue with the execution of a di f ferent action, or training activity, with respect to the performed action, or with respect to the predefined training activity, based on the evaluation score assigned to the latter .

[0102] In more detail , the electronic processing resources 2 are configured to determine and cause the display (block 14 ) of a graphic representation in order that the student 5 continues with the execution of a di f ferent action, or training activity, with respect to the performed action or with respect to the predefined training activity i f it is determined (block 101 ) that the evaluation score is suf ficient to continue ; in particular, i f the evaluation score is higher than a predefined evaluation threshold .

[0103] The electronic processing resources 2 are configured to determine and cause the display (block 14 ) of a graphic representation in order that the student 5 tries again or repeats the execution of the performed action or at least part of the predefined training activity i f it is determined that the evaluation score is insuf ficient to continue ; in particular, i f the evaluation score is below, or possibly equal to , the predefined evaluation threshold or a di f ferent evaluation threshold that de fines the minimum acceptable evaluation score .

[0104] In particular, i f the evaluation score is determined to be insuf ficient to continue , the electronic processing resources 2 , in particular a supervisor module 10 , are configured to cause the displaying (block 14 ) of a debriefing graphic teaching representation (block 102 ) to guide at least partly the student 5 in performing one or more actions , and to allow the student 5 to try again and to repeat the execution of at least part of the predefined training activity .

[0105] According to an aspect of the present invention, the electronic processing resources 2 are configured to allow the student 5 to carry out at least part of the predefined training activity without displaying (block 103 ) the graphic teaching representation i f a predefined condition for deactivating the suggestions is satis fied; preferably, i f the evaluation score is higher than, or possibly equal to , the predefined evaluation threshold that de fines the minimum acceptable evaluation score .

[0106] Optionally, the predefined condition for deactivating the suggestions is satis fied i f the student 5 has repeated part of the predefined activity a number of times greater than a predefined maximum value or i f the student 5 has obtained progressively increasing evaluation scores for a number of previous attempts , expediently predefined at the design stage .

[0107] Furthermore , the electronic processing resources 2 are configured to determine (block 90 ) whether an action performed by the virtual avatar 6 of the student 5 in the rendered virtual working environment 7 is , or is not , a killer action or an action harmful for the predefined training activity; in particular, based on the associated evaluation score , or based on one or more conditions , or rules , prede fined to identi fy (block 90 ) killer actions .

[0108] In particular, the electronic processing resources 2 are configured to determine (block 90 ) whether an action performed by the virtual avatar 6 of the student 5 is , or is not , a killer action via the evaluation model 9 or via the simulation model 8 ; in particular, via the evaluation model 9 and possibly i f it is determined that one or more predefined conditions for identi fying (block 90 ) killer actions are satis fied .

[0109] According to an aspect of the present invention, the simulation model 8 is trained with labels comprising values 13 (by way of example , a Boolean value ) indicative of the fact that an associated action is or is not a killer action, in order to be configured to provide an output indicative of the fact that an action is or is not a killer action . According to said aspect of the present invention, the electronic processing resources 2 are configured to determine (block 90 ) whether an action performed by the virtual avatar 6 of the student 5 is , or is not , a killer action by means of the simulation model 8 trained via supervised learning on a set of training data 11 comprising said labels .

[0110] The electronic processing resources 2 are further designed to cause the display (block 14 ) , by means of the electronic display or proj ection resources 3, of a graphic representation indicative of the need to , or which requires the student 5 to , repeat (block 91 ) the execution of the predefined training activity, at least partially accomplished, i f it is determined (block 90 ) that said action is a killer action . Preferably, i f it is determined (block 90 ) that the performed action is a killer action, said graphic representation requires the student 5 to repeat (block 91 ) the execution of the predefined training activity .

[0111] In detail , the student 5 has accomplished a harmful , or killer, action and must perform the entire activity again since he / she has committed an error which is intended be avoided and not repeated in the future ; in particular, said error is an index of the need to consolidate the basic knowledge of the student 5.

[0112] Otherwise , i f it is determined (block 90 ) that the action performed is not a killer action, the electronic processing resources 2 are configured to activate (block 100 ) , or invoke , the supervisor module 10 designed to determine (block 101 ) whether a predefined learning progress condition of the student 5 is satis fied; in particular, i f the evaluation score is higher than, or possibly equal to , the predefined evaluation threshold that defines the minimum acceptable evaluation score .

[0113] Furthermore , the electronic processing resources 2 are configured to determine a number of actions to be performed in order to accompl ish the training activity, to be taught to the student 5 by means of the teaching representation, based on a learning progress stage of the student 5 determined on the basis of the evaluation scores assigned to the actions performed in the rendered virtual working environment 7 . In further detail , the electronic processing resources 2 are configured to determine the actions to be performed in order to accomplish the training activity, to be taught to the student 5 , based on the learning progress stage of the student 5 determined and based on the simulation model 8 . In particular, the electronic processing resources 2 are configured to determine a number of actions to be performed, in order to accomplish the training activity, which are appropriate for the learning progress stage of the student 5 , namely for the ability level demonstrated in execution of the preceding training activities . In particular, the electronic processing resources 2 are further designed to store the training activities accomplished by the student 5 , both previously and at the current time . In further detai l , the electronic processing resources 2 are configured to determine the learning progress stage of the student 5 , taking account of the activities previously accomplished, due to the fact that the simulation model 8 comprises one or several LSTM neural networks designed, or trained, to capture long-range dependencies in sequences . Furthermore , according to an aspect of the present invention, the electronic processing resources 2 are designed to store the actions and / or sequences of actions commanded by the student 5 and performed in the virtual working environment 7 and the evaluation scores associated with them .

[0114] In particular, said simulation model 8 is designed to determine the action or the sequence of actions to be performed in order to accomplish the predefined training activity, based on the evaluation scores associated with the actions received in input and performed by said student 5 . In fact , since it is designed to determine the action or sequence of actions to be performed based on the as sociation between evaluation scores and the actions or sequences of actions in input , the simulation model 8 is able to determine ( in detail select ) said action or sequence of actions to be performed by setting or aligning itsel f according to the preparation of the student 5 de fined by the evaluation scores associated with the actions already performed by the student 5 . According to an aspect o f the present invention, the simulation model 8 is trained to select the action or sequence of actions to be performed so that it has an evaluation score in line with the evaluation scores associated with the actions already performed by the student 5 .

[0115] In detail , said s imulation model 8 is designed to determine the action or sequence of actions to be performed in order to accomplish the predefined training activity, based on the learning progress stage of the student 5 . Possibly, the simulation model 8 is designed to determine , in detail compute , said learning progress stage of the student 5 based on the evaluation scores assigned to the actions performed in the rendered virtual working environment 7 in order to accomplish the predefined training activity . In detail , the neural simulation network of the simulation model 8 is configured to determine the action or sequence of actions to be performed in order to accomplish the predefined training activity based on the learning progress stage of the student 5 since it is trained to determine relationships and / or dependencies in the sequences of actions and in their associated evaluation scores . In detail , in order to determine the action or sequence of actions to be performed in order to accompli sh the predefined training activity, the simulation model 8 is designed to determine the evaluation score of an action or sequence of actions which the student 5 could perform based on the sequence of actions received in input and based on the learning progress stage o f the student 5 determined on the basis of said sequence of actions received in input .

[0116] According to an aspect of the present invention, in order to propose an action or sequence of actions appropriate to the learning progress stage of the student 5 , the electronic processing resources 2 are configured to communicate with a database containing the data of a plurality of students previously trained on the same training activity . This database contains preferably at least the sequence of actions performed for the execution of an activity and the score associated with each of them . Furthermore , in particular, the database contains an identi fication of each student and an identi fication of the activities performed . Possibly, the electronic processing resources 2 are configured to update said database every time a student trains , storing in the latter the actions performed by said student in training and the evaluation scores associated with them .

[0117] According to said aspect of the present invention, the electronic processing resources 2 are designed to perform a clustering algorithm to receive in input , in detail from said database , di f ferent sequences of actions ( or time series ) of the preceding activities of each student , and to create clusters for similar students in relation to the same activities performed and the same evaluation scores . In particular, said clustering algorithm is designed to cause , when executed, the electronic processing resources 2 to become configured to create clusters based on ability and / or skills shared by a plurality of students . In particular, said clustering algorithm is designed to generate di f ferent groups of students , with similar abilities and / or skills , based on the sequences of actions of the preceding evaluations of the activities of each student received in input . Said clustering algorithm is preferably a non-supervised learning algorithm .

[0118] Furthermore , the electronic processing resources 2 , executing the clustering algorithm, are preferably configured to label each group of students ( or cluster ) generated via the execution of said clustering algorithm . Possibly, the electronic processing resources 2 are configured to compute and associate the learning progress stage of a group of students generated on the basis of the evaluation scores associated with the sequences of actions of the students grouped in said group of students . For example , a learning progress stage of a group of students could be an evaluation score , or a range of evaluation scores , computed on the basis of the evaluation scores associated with the sequences of actions of the students grouped in said group of students . Furthermore , for each group of students generated, the electronic processing resources 2 are preferably configured to label each group of students ( or cluster ) generated based on the learning progress stage associated with said group . By way of example , the electronic processing resources 2 are configured to label each group of students ( or cluster ) generated with a textual content indicative of the learning progress stage associated with said group .

[0119] In particular, this provides a distribution function relative to all the scores that define the elements of the group, so as to obtain a reference standard .

[0120] By way of non-limiting example , the clustering algorithm used is the DBScan or, alternatively, the TimeSeriesKMeans . The TimeSeriesKMeans is an algorithm optimi zed to work on numerical data and on time series and, unlike the DBScan, is designed to receive in input a predefined number of groups to perform the clustering . For example , considering three groups , the evaluation associated and defining the corresponding label of the groups could be as follows : 'poor, average , good' . The DBScan, on the other hand, is an algorithm that does not envisage the number of groups in input , but is configured to collect and organi ze the data as a function of the number of groups which it determines to be most expedient . In this way the granularity of the evaluation and therefore the consequent labelling of the groups is defined based on the number of groups obtained . This determines an alternative which is even more guided by the type of data collected . In this way, since it is continuously produced in relation to the standard obtained from the distribution, each evaluation score is related to the level of all the other scores of students with the same technical abilities .

[0121] In detail , the electronic processing resources 2 , in detail by executing the clustering algorithm, are designed to associate the student 5 with a membership group ( groups obtained by means of the clustering algorithm) based on the scores associated with the actions or sequences of actions performed by said student 5. In further detail , the electronic processing resources 2 are designed to select a group of the groups of students generated by means of the clustering algorithm based on said scores associated with the actions or sequences of actions performed by said student 5 . Furthermore , said electronic processing resources 2 are designed to associate the student 5 with the selected group ; in detail , to include said student 5 ( or an identi fication of the latter ) in said selected group . Subsequently, the electronic processing resources 2 are possibly designed to determine the evaluation score of an action or sequence of actions of said student 5 by making a comparison with the previously obtained distribution function of the scores . In particular, said distribution function will be the more precise and accurate the more data are accumulated on the students in the database . In particular, the electronic processing resources 2 are designed to determine the evaluation score of an action of the student 5 based on the group associated with said student 5 , expediently on the basis of the standard reference distribution identi fied .

[0122] Optionally, the simulation model 8 is further designed to determine an index of di f ficulty ( or complexity) of an action or sequence of actions , indicative of the preparation necessary in order to perform said action or sequence of actions on the basis of the latter . In further detail , in order to determine the action or sequence of actions to be performed by the student 5, the simulation model 8 is designed to determine , in detail compute , an evaluation score of an action or sequence of actions based on the index of di f ficulty of said action or sequence of actions and based on the learning progress stage of the student 5. For example , an evaluation score is relatively high for an action or sequence of actions whose index of di f ficulty is relatively low with respect to the learning progress stage of the student 5 , and it is relatively low for an action or sequence of actions whose di f ficulty index is relatively high with respect to the learning progress stage of the student 5 . In this way, the electronic processing resources 2 , and in detail the simulation model 8 , are able to assign to the student 5 actions according to his / her learning progress stage , prioritising the actions that have ( or are associated with) an index of di f ficulty aligned with his / her learning progress stage .

[0123] Optionally, said simulation model 8 is designed to output the action or sequence of actions ( in detail modelled by a time series ) which the virtual avatar 6 of the student 5 could perform also based on further and di f ferent data indicative of the knowledge and / or skills developed by the student during his / her training course .

[0124] Furthermore , the electronic processing resources 2 are configured to cause the display (block 14 ) , by means of the electronic display or proj ection resources 3, of the teaching representation to guide the student 5 in learning the actions determined and to be performed in order to accomplish the training activity . In particular, upon completion of each training activity, the electronic processing resources 2 are configured to cause the display (block 14 ) , by means of the electronic display or proj ection resources 3, of the teaching representation to guide the student 5 to continue in a di fferent and subsequent training activity of the predefined training course by means of a sequence of actions , determined on the basis of the learning progress stage , appropriate for the ability level demonstrated by the student 5 up to that moment .

[0125] By way of non-limiting example , a situation could be observed in which the pilot student 5 is beginning to become familiar with a system ( for example the electrical system) of the aircraft , and an anomaly occurs to be solved on that system . I f the virtual avatar 6 of the pilot 5 intervenes by actuating a sensor assigned for example to activation of the undercarriage system, then the training system for human operators 1 provided would propose a sequence of actions that enable the student to become familiar with the arrangement of the sensors throughout the cockpit , relative to the electrical system . Namely, again according to said example , the training system for human operators 1 will show the various sensors that can intervene to solve the situation, and wait for a new input from the virtual avatar 6 ; based on the input received, it will propose a sequence of actions to teach the pilot student 5 how to perfect his / her activity to solve the problem ( for example by showing said sequence of actions in an order of priority) . Alternatively, again according to said example , the training system for human operators 1 will show a sequence of actions , following a "poorer" action of the pilot student 5 , which will enable him / her to become aware of the entire arrangement of the cockpit sensors , viewing each sensor and the correlation with any further systems of the aircraft . By way of non-limiting example , the training system for human operators 1 is further configured to cause the displaying in the virtual working environment 7 of warnings , or pop ups , when a virtual avatar 6 presses the ENGINE FAILURE button; in which an overlapping of the image of the engine is displayed . In detail , the training system for human operators 1 could propose a sequence of actions connectable to the first sequence of actions proposed, having di f ferent di f ficulty with respect to the f irst sequence ( for example , a higher level of di f ficulty) .

[0126] On the basis of the above description, the advantages of the present invention are evident .

[0127] In particular, the Applicant has observed that the present invention allows the training of aircraft pilots to be standardi zed to the highest possible level .

[0128] Furthermore , the present training software for human operators 1A allows a subj ective evaluation of a human instructor to be entirely replaced by an obj ective evaluation .

[0129] Furthermore , the Applicant has observed that the present invention allows indications to be provided to a student 5 , appropriate for the ability level demonstrated, on how to continue after the completion of a training activity and therefore after achieving a training obj ective .

[0130] Furthermore , the Applicant has observed that the present invention allows a student 5 to be guided in detail in performing a series of speci fic actions in order to achieve a training obj ective . In particular, the Applicant has observed that the present invention allows determination and teaching of the best speci fic actions which the student 5 could perform, in the situation ( in particular, defined by the actions performed at times previous to the current time ) in which he / she finds himsel f / hersel f , in order to achieve a predef ined training ob ective .

Claims

CLAIMS1. Training software for human operators (1A) , for example for a pilot student (5) , to teach him or her to accomplish a training activity, for example a piloting training activity; the training software for human operators (1A) is storable in and executable by electronic processing resources (2) and configured to cause, when executed, said electronic processing resources (2) to become configured to:- cause the display (block 14) , by means of electronic display or projection resources (3) , of a rendered virtual work environment (7) corresponding to a real work environment in which the student (5) will have to operate;- receive (block 14) from an electronic controller (4) , operable by the student (5) , a number of commands imparted by the student (5) and indicative of one or different actions to be performed by a virtual avatar (6) of the student (5) in the rendered virtual working environment (7) ;- cause the display (block 14) , by means of the electronic display or projection resources (3) and in response to, and on the basis of, the commands received from the electronic controller (4) , of the execution of the actions associated with these commands and performed by the virtual avatar (6) of the student (5) in the rendered virtual working environment (7) ;- implement a simulation model (8) configured to determine an action, and / or a sequence of actions, to be performed in order to accomplish a predefined training activity, in particular a piloting one, based on a plurality (80) of actions and / or of sequences of actions and based on evaluation scores (81) associated with them; wherein, the simulation model (8) comprises at least one neural simulation network trained on a plurality of sample actions, and / or sequences of sample actions (80) , and evaluation scores (81) associated with them, to determine an action, and / or a sequence of actions, to be performed based on the actions performed by a virtual avatar (6) of the student (5) and received in input; and wherein theevaluation scores (81) are indicative of the contribution of such actions, or sequences of actions, (80) to the accomplishment of the predefined training activity in the rendered virtual working environment (7) ;- determine a number of actions to be performed, in order to accomplish a training activity to be learned, based on one or more actions performed by the virtual avatar (6) of the student (5) in the rendered virtual working environment (7) , and based on the simulation model (8) implemented; and- cause the display (block 14) , by means of the electronic display or projection resources (3) , of a teaching representation configured to guide the student (5) in learning the determined actions to be performed in order to accomplish the training activity to be learned in the rendered virtual working environment (7) .

2. The training software for human operators (1A) according to claim 1, wherein the simulation model (8) is configured to: receive in input information indicative of the actions performed by the virtual avatar (6) of the student (5) ;- predict an evaluation score for a number of actions, which the virtual avatar (6) of the student (5) could perform, identified based on the performed actions; and- determine and output an action or sequence of actions to be performed based on a number of predicted evaluation scores.

3. The training software for human operators (1A) according to claim 1 or 2, wherein the simulation model (8) is configured to determine an action, and / or a sequence of actions, to be performed in order to accomplish a predefined training activity also based on the predefined training activity.

4. The training software for human operators (1A) according to claim 3, wherein the simulation model (8) , for each of several predefined training activities, comprises a simulation sub-model (8) associated with it; furthermore, each of the simulation sub-models is configured to determine an action, and / or a sequence of actions to be performed, to accomplish the predefined training activity associated with it , starting from the accomplished actions received in input .

5. The training software for human operators ( 1A) according to any one of the preceding claims , wherein the simulation neural network is a recurrent neural network (RNN) trained to capture one or several relationships in a sequence of actions , comprising the actions performed and one or more actions to be performed in sequence , in order to predict the evaluation score of said sequence of actions .

6. The training software for human operators ( 1A) according to any one of the preceding claims , and designed to cause , when executed, the electronic processing resources ( 2 ) to become further configured to :- implement or receive an evaluation model ( 9 ) configured to assign an evaluation score to an action, and / or a sequence of actions , to be performed in order to accomplish a predefined training activity based on a plurality of actions and / or of sequences of actions ( 80 ) and based on evaluation scores ( 81 ) associated with them;- determine and assign an evaluation score to one or more of the actions performed by the virtual avatar ( 6 ) of the student (5 ) , in order to accomplish at least partially the predefined training activity, on the basis of the evaluation model ( 9 ) ; and- cause the display (block 14 ) , by means of the electronic display or proj ection resources ( 3 ) , of a teaching representation, indicative of a learning progres s stage of the student ( 5 ) in relation to the predefined training activity, based on a number of evaluation scores assigned to one or more actions performed in the rendered virtual work environment ( 7 ) in order to accomplish the predefined training activity .7 . The training software for human operators ( 1A) according toclaim 6, and designed to cause, when executed, the electronic processing resources (2) to become configured to:- cause the display (block 14) , by means of the electronic display or projection resources (3) , of a graphic representation indicative of the need to, or requiring the student (5) to, repeat the execution of the predefined training activity, or of at least part of the predefined training activity, or to continue with the execution of a different action, or training activity, with respect to the action performed, or to the predefined training activity, on the basis of the evaluation score assigned to the performed action.

8. The training software for human operators (1A) according to any preceding claim, and designed to cause, when executed, the electronic processing resources (2) to become configured to:- determine (block 90) whether an action performed by the virtual avatar (6) of the student (5) in the rendered virtual working environment (7) is, or is not, a killer action or a harmful action for the predefined training activity; and- cause the display (block 14) , by means of the electronic display or projection resources (3) , of a graphic representation indicative of the need to, or requiring the student (5) to, repeat (block 91) the execution of the predefined training activity, at least partially accomplished, if it is determined (block 90) that this action is a killer action.

9. The training software for human operators (1A) according to any preceding claims, and designed to cause, when executed, the electronic processing resources (2) to become configured to determine said actions to be performed, in order to accomplish a training activity to be learned, based on a learning progress stage of the student (5) determined on the basis of the evaluation scores assigned to the actions performed in the rendered virtual working environment (7) .

10. A training system for human operators, for example for apilot student (5) , to teach him or her to accomplish a training activity, for example a piloting training activity; the training system for human operators comprising:- electronic display or projection resources (3) designed to render, or to display (block 14) , a rendered virtual work environment (7) corresponding to a real work environment in which a student (5) will have to operate;- an electronic controller (4) operable by a student (5) to impart a number of commands indicative of one or different actions to be performed by a virtual avatar (6) of the student(5) in the rendered virtual working environment (7) ; and- electronic processing resources (2) storing, and designed to execute, the training software for human operators (1A) of any one of the preceding claims.

Images