EP4384358A1 - Verfahren und system zur analyse eines betriebs eines roboters - Google Patents
Verfahren und system zur analyse eines betriebs eines robotersInfo
- Publication number
- EP4384358A1 EP4384358A1 EP22760708.2A EP22760708A EP4384358A1 EP 4384358 A1 EP4384358 A1 EP 4384358A1 EP 22760708 A EP22760708 A EP 22760708A EP 4384358 A1 EP4384358 A1 EP 4384358A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- robot
- autoencoder
- data set
- patterns
- time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Program-controlled manipulators
- B25J9/16—Program controls
- B25J9/1628—Program controls characterised by the control loop
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Program-controlled manipulators
- B25J9/16—Program controls
- B25J9/1674—Program controls characterised by safety, monitoring, diagnostic
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/33—Director till display
- G05B2219/33296—ANN for diagnostic, monitoring
Definitions
- the present invention relates to a method and system for analyzing an operation of a robot and a computer program or
- An object of an embodiment of the present invention is to analyze an operation of a robot or to identify significant patterns or sections in robot data over time.
- a robot anomaly or an event for example an environmental contact
- a time course can be classified, for example a specific robot movement within a work process can be recognized again.
- the operation of a robot can thus be analyzed, monitored and/or modified, for example by checking detected robot anomalies or by carrying out maintenance, in particular predictive, on the basis of detected robot anomalies, an operating sequence based on a detected event, for example environmental contact, modified and/or monitored for such event, or the like.
- Claims 9, 10 represent a system or computer program or
- a method for analyzing an operation of a robot comprises performing a training phase comprising the steps:
- this artificial neural network is a neural network that uses this artificial neural network to train an artificial neural network. According to an embodiment of the present invention, this artificial neural network
- a second autoencoder with an encoder that maps the time course patterns and their activation to pattern groups and in a development their activation, and a decoder that reconstructs the time course patterns and their activation with these pattern groups and in the development of their activation; on.
- the method has a monitoring phase carried out, which has the following steps:
- This identification can represent an analysis(s) of the operation of this robot, for which the second data set has been obtained, in the sense of the present invention.
- a (further) analysis can be carried out on the basis of this identified at least one pattern group, in one embodiment automatically and/or manually or by a user.
- a time course can include, in particular be, a time series.
- the first autoencoder which can also be referred to as a low-level autoencoder, has at least one variational autoencoder.
- the second autoencoder can also be referred to as a high level autoencoder.
- the encoder of the second autoencoder has at least one attention-based artificial neural network (attention-based neural network), in particular at least one multihead attention block. Additionally or alternatively, in one embodiment, the decoder of the second autoencoder has at least one capsule neural network.
- attention-based artificial neural network attention-based neural network
- multihead attention block at least one multihead attention block.
- the decoder of the second autoencoder has at least one capsule neural network.
- such autoencoders can be used to identify pattern groups in robot data over time in a particularly efficient manner by machine learning.
- the state parameter depends on at least one position and/or orientation of a robot-fixed reference, in particular an end effector and/or one or more axes or drives, and/or at least one axis load of the (respective) robot he/she indicates this position(s) and/or orientation(s) and/or axle load(s) and/or changes in this position(s) and/or orientation(s) and/or axle load(s) over time.
- the state parameter is recorded by means of one or more sensors, in one embodiment on the robot or on the robot.
- the identified pattern group is marked in the second data record, in one embodiment optically or visually.
- a user can concentrate on particularly relevant sections in the course of time.
- a robot anomaly is detected based on the identified pattern set.
- maintenance in particular predictive maintenance, of the robot can be improved.
- an event for example an, in particular unforeseen or planned, environmental contact is detected.
- operation of the robot can be monitored particularly advantageously, for example for unforeseen collisions, and/or, for example as a function of planned environmental contact, modified, for example, an action can be carried out depending on a scheduled contact with the environment.
- the time history of the second data record is classified, in one embodiment as to whether the robot has performed a specific action and/or specific boundary conditions, in particular environmental conditions, have existed.
- the corresponding time course can then be used for machine learning, for example on the basis of this (automatic) classification.
- an identified pattern group within a dataset containing one or more robot state parameter time histories can be used to advantage to analyse, monitor and/or modify an operation of a robot.
- operation of the first or second robot is based on the identified pattern group, in one embodiment on the basis of the pattern group marked in the second data set and/or the detected robot anomaly and/or the detected event and/or the classified time course of the second data set, analysed, monitored and/or modified.
- the identification can already represent an analysis(s) of the operation of the robot within the meaning of the present invention or, in one embodiment, a (further) analysis can be carried out on the basis of this identified at least one pattern group.
- a system for analyzing the operation of a robot in particular in terms of hardware and/or software, in particular in terms of programming, is set up to carry out a method described here and/or has:
- an artificial neural network which has a first autoencoder with an encoder, which maps the first data set to time history patterns and their activation, and a decoder that reconstructs the first data set with these timing patterns; and a second autoencoder with an encoder that reads the
- the means for carrying out a training phase has:
- the artificial neural network which has the first autoencoder and the second autoencoder; and or
- the means for carrying out a monitoring phase has:
- system or its means(s) has:
- a system and/or a means within the meaning of the present invention can be designed in terms of hardware and/or software, in particular at least one, in particular digital, processing unit, in particular microprocessor unit ( CPU), graphics card (GPU) or the like, and / or have one or more programs or program modules.
- the processing unit can be designed to process commands that are implemented as a program stored in a memory system, to detect input signals from a data bus and/or to output output signals to a data bus.
- a storage system can have one or more, in particular different, storage media, in particular optical, magnetic, solid-state and/or other non-volatile media.
- the program may be of such a nature that it embodies or is capable of carrying out the methods described herein, so that the processing unit can carry out the steps of such methods and thereby in particular identify the pattern group(s) and/or analyze, monitor and/or an operation of the robot or can modify.
- a computer program product can have, in particular, be a, in particular, computer-readable and/or non-volatile storage medium for storing a program or instructions or with a program or with instructions stored thereon.
- execution of this program or these instructions by a system, in particular a computer or an arrangement of multiple computers causes the system, in particular the computer or computers, to carry out a method or one or more of its steps described here , or the program or the instructions for this are set up.
- one or more, in particular all, steps of the method are carried out fully or partially automatically, in particular by the system or its means.
- the system includes the robot.
- the monitoring phase is carried out online, in one embodiment during a work process of the first or second robot.
- the second data record results from a work process of the first or second robot.
- the artificial neural network tries to generate the same output as the time profile of the first data set obtained and at the same time to extract recurring and discriminative partial sequences from the time profile. Based on the presence and localization of various recurring and discriminative partial sequences, in one embodiment a time course is then clustered into a predetermined number of groups.
- an artificial neural network with a bottleneck for reconstructing the original input signal is referred to as an autoencoder.
- the first or low-level autoencoder attempts to extract distinguishing but also recurring time course subsequences during the reconstruction of the input time courses.
- the first autoencoder can be designed in an implementation based on the principles described in Kirschbaum, E. et al. (2019): “LeMoNADe: Learned Motif and Neuronal Assembly Detection in calcium imaging videos", International Conference on Learning Representations. Accordingly, additional reference is made to this article and its content is made part of the present disclosure in its entirety.
- the first data set has at least one n-dimensional time series xe HV xd with T journals and d status parameters, which are assigned to a journal and are in particular stored for this purpose.
- each sample in the dataset is an additive mixture of M repeating (time course) patterns of maximum temporal length F.
- a latent random variable z t m e ⁇ 0,1 ⁇ represents an occurrence or Activation encoding for the m-ie pattern.
- a Bernoulli distribution is used to model the latent random variables.
- the loss function for training includes a reconstruction error E z ⁇ 0 [logp e (x
- the second or high-level autoencoder tries to cluster an input time series into different groups.
- An input time series can be assumed to be composed of a plurality of time series subsequences or timing patterns extracted by the first or low-level autoencoder. Irrespective of the numbering of low-level time series segments, the second or high-level autoencoder tries to recognize in one execution whether a new subsequence is present in the current input time series but is missing in all past input time series. If this is the case, the second or high-level autoencoder classifies this time series as a new class.
- the second high-level autoencoder performs clustering based on the assembly of low-level time-series subsequences.
- a set transformer is an attention-based neural network to model interactions between sets of data.
- it consists of one or more Multihead Attention Blocks (MAB) with learnable parameters co, which can be defined as follows (in this regard, reference is also made to Lee, J. et al. (2019), “Set Transformer: A Framework for Attentionbased Permutation-Invariant Neural Networks", Proc, of the 36th Int. Conference on Machine Learning, pp. 3744-3753 and the contents of this article are made a part of the present disclosure in their entirety): Let there be two d - dimensional vector sets: X, Y e IR. n xd ,
- LayerNorm stands for layer normalization, which normalizes the activities of the neurons of a layer. Layer normalization, in one embodiment, can advantageously stabilize the dynamics of hidden neuron states and thereby training time can be reduced.
- Multihead(X, Y, Y; a> ) is also known as multi-head attention and is a way to calculate relationships between neurons in an artificial neural network. In one implementation, this includes three variables: "query”, "key” and "values".
- a d - dimensional vector set X is used as a "query” for attention, in one embodiment a list of detected low-level time-series sub-sequences.
- the output/"values" Y after an attention operation should be a composition of detected low-level time series patterns and the "keys" of this attention operation used are in one embodiment some, preferably all, possible combinations of one, preferably all , low-level time-series subsequences that are unexposed in a training dataset that results in the “query” sentence.
- This type of attention operation is also referred to as self-attention.
- Multi-head Attention projects, in one execution, query(Q)/key(A')/values (V), onto lower-dimensional h common subspaces, ie dq , dq , d"-dimensional vectors.
- An attention operation (Att( -; (o 7 )) is applied to each of these h projections
- Multi-head Attention allows it in one Making the model access information from different representations or subspaces together.
- the output is a linear transformation of the concatenation of all subspace attention outputs.
- the positional information is injected by adding the positional encoding to the projected common space of the query and key vectors.
- the output of the set transformer described above represents a list of candidate combinations of determined low-level time-series subsequences or patterns in each input and/or corresponds to a list of proposed complex candidate patterns or pattern groups.
- the record transformer is used as an encoder for the second or high-level autoencoder. Additionally or alternatively, in one embodiment, a capsule neural network or capsule module is used as a decoder for the second or high-level autoencoder.
- the product is a capsule Operation a vectorial value that provides the activation of certain features and additionally an embedding to describe the activation features, e.g. a localization.
- the Capsule Neural Network may be designed based on the principles set forth in Kosiorek, A. et al. (2019), “Stacked Capsule Autoencoders", Advances in Neural Information Processing Systems, pp. 15512-15522. Accordingly, additional reference is made to this article and its content is made part of the present disclosure in its entirety.
- a complex time series pattern or pattern group can be assumed to have the same length as input time series but with instance-dependent transformation.
- a time series of the same type can look visually different due to different sampling rates.
- a capsule represents such a typical time series of a specific kind, and the associated encapsulation describes a warping variation of that standard.
- localizations of the characteristic time series subsequences contained in it can be derived from a complex time series pattern.
- the latent space of the second or high-level autoencoder has clustered groups or pattern groups of determined characteristic low-level time series segments or time course patterns.
- the latent space is represented in the form of capsules, and each represents a clustered set and its associated embeddings.
- the goal of the second or high-level autoencoder is to train the clustered groups that represent complex patterns of input data length.
- the derived locations of the clustered group across the positions of the low-level time-series subsegments match those learned from the first and low-level autoencoder, respectively.
- time series patterns are extracted without human supervision and extracted patterns are displayed other unknown new robot state parameter time histories generalized. Additionally or alternatively, in one embodiment, matching a learned pattern of one, in particular the first, dataset in a new, in particular the second, dataset may associate this new dataset without information with a known dataset with labeled details. In this way, in one embodiment, the amount of data that is available for monitoring and, in particular, predictive maintenance can be increased.
- the present invention includes one or more of the following applications:
- the first or second data set comprises time histories from different situations in a robot application.
- a timeline from the same scenario can be considered the same type.
- certain segments of a live data series are marked. These salient segments are subsequences from the given data set that differ most from previously recorded sets of other situations.
- a user can focus on these specific segments to perform analysis or diagnosis. For example, user can check motor torque or motor current within the time interval as indicated by the characteristic subsequences.
- Event detection A spike in force absorption on a robotic end effector can represent a collision. A continuously increasing level of force may indicate a planned contact event.
- Comparison of old data without manual labeling and new, labeled data In one embodiment, a known pattern that is present in the labeled data is searched for in the unlabeled data. In one embodiment, if the same time-series subsequence reappears in the old data, the robot is determined to have operated under similar conditions. Consequently, a user can transfer his knowledge based on a subsequence from the new labeled data to analyze old, unlabeled data with less information. 4.
- Machine learning to learn recurring time-series primitives such as pulses, trapezoidal segments, square waves, or the like. In one embodiment, transformations are performed on these primitives to locate variations thereof in a live dataset. This function is particularly useful for estimating or observing hidden physical variables of the robot state, such as axis friction, based on time series segments with certain properties.
- a user is provided with representative templates for each time-series data group.
- Activation within the meaning of the present invention can in particular include localization and/or embedding within the meaning of the present invention and vice versa.
- Fig. 1 a system according to an embodiment of the present invention.
- Figure 2 a method according to an embodiment of the present invention.
- Fig. 1 shows a system according to an embodiment of the present invention
- Fig. 2 shows a method according to an embodiment of the present invention.
- a first data set with a time profile xi(t) of a state parameter of a robot 1 is obtained (FIG. 2: step S10).
- the status parameter is, for example, a drive torque control error.
- an artificial neural network is trained, which has a first autoencoder with an encoder 11, which maps the first data set to timing patterns and their activation, and a decoder 12, which reconstructs the first data set with these timing patterns. and a second autoencoder with an encoder 21, which maps the time course patterns and their activation to pattern groups, and a decoder 22, which reconstructs the time course patterns and their activation with these pattern groups.
- a second data set X2(t) with a time profile of the status parameter of the robot 1 is obtained (FIG. 2: step S30).
- one of the pattern groups of the trained second autoencoder is identified within the second data set and marked in the second data set.
- This pattern group is indicated in Fig. 1 by a hatched rectangle with a dashed edge: first, time progression patterns and their activation, which are initially randomly occupied in the latent space 13 of the first autoencode, are learned or the first autoencoder is trained in such a way that the time progression xi(t) reconstructed as well as possible.
- timing patterns and activations learned in this way are entered into the second autoencoder.
- all possible combinations of time course patterns and activations are initially occupied on a random basis.
- the second autoencoder is now trained in such a way that it reconstructs the time course patterns and activations from the first autoencoder (“low-level time series subsequences”) as well as possible.
- time profile pattern 3 shown in FIG. 1 occurs only once at a specific point in the time profile xi(t), which is indicated in FIG. 1 by a hatched rectangle with a dashed edge.
- This time profile pattern 3 is now searched for or identified and marked in the new time profile X2(t), which is indicated in FIG. 1 by a hatched rectangle with a dashed edge.
- the time profile X2(t) is recorded during a work process of the robot 1, for example.
- a significant period of time can be recognized in the course of time X2(t) or work process, for example a collision, contact with the environment, a robot anomaly or the like.
- step S40 the operation of the robot can be analysed, monitored and/or modified in step S40 on the basis of this identified time course pattern, for example if a (specific) robot anomaly occurs, predictive maintenance can be initiated accordingly.
- a history of the drive torque control error due to the environmental contact can be further analyzed or the like.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Manipulator (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| DE102021208769.8A DE102021208769B8 (de) | 2021-08-11 | 2021-08-11 | Verfahren und System zur Analyse eines Betriebs eines Roboters |
| PCT/EP2022/071511 WO2023016837A1 (de) | 2021-08-11 | 2022-08-01 | Verfahren und system zur analyse eines betriebs eines roboters |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP4384358A1 true EP4384358A1 (de) | 2024-06-19 |
Family
ID=83081568
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP22760708.2A Pending EP4384358A1 (de) | 2021-08-11 | 2022-08-01 | Verfahren und system zur analyse eines betriebs eines roboters |
Country Status (5)
| Country | Link |
|---|---|
| US (1) | US20240428051A1 (de) |
| EP (1) | EP4384358A1 (de) |
| CN (1) | CN118103178A (de) |
| DE (1) | DE102021208769B8 (de) |
| WO (1) | WO2023016837A1 (de) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20240112016A1 (en) * | 2022-09-30 | 2024-04-04 | Falkonry Inc. | Scalable, multi-modal, multivariate deep learning predictor for time series data |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102014226787B3 (de) | 2014-12-22 | 2016-03-17 | Kuka Roboter Gmbh | Sicherer Roboter mit Bahnfortschrittsvariablen |
| DE102017125330B3 (de) | 2017-10-27 | 2019-01-10 | Kuka Deutschland Gmbh | Virtuelle Funktionsauslöser |
| DE102018203234A1 (de) | 2018-03-05 | 2019-09-05 | Kuka Deutschland Gmbh | Vorausschauende Beurteilung von Robotern |
| DE102019214009B3 (de) | 2019-09-13 | 2020-12-10 | Kuka Deutschland Gmbh | Analyse von Sensormessdaten eines Roboters oder Manipulators |
| DE102019220574A1 (de) | 2019-12-27 | 2021-07-01 | Robert Bosch Gmbh | Verfahren und Vorrichtung zum Testen einer Maschine |
| DE102020210537A1 (de) | 2020-08-19 | 2022-02-24 | Kuka Deutschland Gmbh | Verfahren und System zum Handhaben einer Lastanordnung mit einem Robotergreifer |
-
2021
- 2021-08-11 DE DE102021208769.8A patent/DE102021208769B8/de active Active
-
2022
- 2022-08-01 CN CN202280068758.1A patent/CN118103178A/zh active Pending
- 2022-08-01 US US18/683,017 patent/US20240428051A1/en active Pending
- 2022-08-01 WO PCT/EP2022/071511 patent/WO2023016837A1/de not_active Ceased
- 2022-08-01 EP EP22760708.2A patent/EP4384358A1/de active Pending
Also Published As
| Publication number | Publication date |
|---|---|
| US20240428051A1 (en) | 2024-12-26 |
| DE102021208769B8 (de) | 2023-02-23 |
| DE102021208769B3 (de) | 2023-01-05 |
| CN118103178A (zh) | 2024-05-28 |
| WO2023016837A1 (de) | 2023-02-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhou et al. | Interpreting deep visual representations via network dissection | |
| EP3847578A1 (de) | Verfahren und vorrichtung zur klassifizierung von objekten | |
| DE102018111905A1 (de) | Domänenspezifische Sprache zur Erzeugung rekurrenter neuronaler Netzarchitekturen | |
| DE112019000093T5 (de) | Diskriminierungsvorrichtung und Maschinenlernverfahren | |
| DE112017005651T5 (de) | Vorrichtung zur Klassifizierung von Daten | |
| EP3557487B1 (de) | Generieren von validierungsdaten mit generativen kontradiktorischen netzwerken | |
| DE102014108287A1 (de) | Schnelles Erlernen durch Nachahmung von Kraftdrehmoment-Aufgaben durch Roboter | |
| DE102023212504A1 (de) | Systeme und Verfahren zum Training eines Videoobjekt-Detektionsmaschinen-Lernmodells mit einem Rahmen von Lehrer und Schüler | |
| EP3696743A1 (de) | Verfahren und überwachungsvorrichtung zur überwachung eines technischen systems mittels anomalieerkennung | |
| DE102021204040A1 (de) | Verfahren, Vorrichtung und Computerprogramm zur Erstellung von Trainingsdaten im Fahrzeug | |
| DE19636074C2 (de) | Lernfähiges Bildverarbeitungssystem zur Klassierung | |
| EP4384358A1 (de) | Verfahren und system zur analyse eines betriebs eines roboters | |
| DE112021007196T5 (de) | Maschinenlernvorrichtung, klassifizierungsvorrichtung und steuervorrichtung | |
| EP3330818A1 (de) | Verfahren und vorrichtung zur zustandsüberwachung von komponenten einer technischen anlage | |
| DE102019209228A1 (de) | Verfahren und Vorrichtung zum Überprüfen der Robustheit eines künstlichen neuronalen Netzes | |
| DE112021008311T5 (de) | Arbeitsanalysevorrichtung | |
| DE102022107831A1 (de) | Neuronale Netzwerkarchitektur für die automatisierte Teileprüfung | |
| DE102018128640A1 (de) | Vorrichtung zur Einschätzung einer Rauscherzeugungsursache | |
| Casacuberta et al. | PCACE: A statistical approach to ranking neurons for CNN interpretability | |
| DE102025112668A1 (de) | System und Verfahren zum Vorhersagen diverser zukünftiger Geometrien mit Diffusionsmodellen | |
| Eum et al. | Region-based conversion of neural activity across sessions | |
| DE102023207212A1 (de) | Verfahren und ein System zum optimierten Trainieren eines Algorithmus des maschinellen Lernens | |
| EP4645162A1 (de) | Verfahren und system zur inspektion oder zum betrieb eines produkts | |
| DE102004018174B4 (de) | Verfahren zur Akquisition von Formen aus Bildern mit Fällen und zum fallbasierten Erkennen von Objekten in digitalen Bildern, Computer-Programm-Produkt und digitales Speichermedium zur Ausführung dieses Verfahrens | |
| EP4636662A1 (de) | Verfahren zum trainieren eines gams |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
| PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
| 17P | Request for examination filed |
Effective date: 20240206 |
|
| AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
| P01 | Opt-out of the competence of the unified patent court (upc) registered |
Free format text: CASE NUMBER: APP_48467/2024 Effective date: 20240823 |
|
| DAV | Request for validation of the european patent (deleted) | ||
| DAX | Request for extension of the european patent (deleted) |