A controller for training a machine to automate lighting control actions and methods thereof

Abstract

A method for training a machine to automate lighting control actions, wherein the method comprises the steps of: controlling one or more lighting devices to render a first light effect based on a first set of control parameters; controlling the one or more lighting devices to render a second light effect based on a second set of control parameters by transitioning from the first light effect to the second light effect over a transition period; receiving feedback from a user during a feedback period; wherein the feedback is associated with the transition if the feedback has been received within a predetermined period of time during the feedback period; and wherein the feedback is associated with the second light effect if the feedback has been received after the predetermined period of time during the feedback period; and wherein the method further comprises training the machine based on the associated feedback.

Description

Technical Field

[0001] This invention relates to a method for training a machine to automate lighting control actions. The invention also relates to a controller, system, and computer program product for training a machine to automate lighting control actions. Background Technology

[0002] Connected lighting refers to a system of one or more lighting devices that are not controlled by (or solely by) traditional wired, electrical switch, or dimmer circuitry, but rather via data communication protocols using wired or more commonly wireless connections (such as wired or wireless networks). These connected lighting networks form what is commonly known as the Internet of Things (IoT), or more specifically, the Internet of Lighting (IoL). Typically, a lighting device, or even individual lights within a lighting device, may be equipped with its own wireless receiver or transceiver to receive lighting control commands from the lighting control device according to wireless network protocols such as Zigbee, Wi-Fi, or Bluetooth.

[0003] These lighting devices in a connected lighting system can be controlled to present different lighting effects, for example, based on user preferences. Advances in machine learning have made it possible to understand which lighting effects a user prefers. Users can provide feedback on their preferences for different lighting effects, and self-learning systems using machine learning can learn the user preferences associated with different lighting effects. The trained model can then be used to automatically control the lighting devices to present the lighting effects preferred by the user.

[0004] US 2017 / 285594A1 discloses a device including a processor, a processor-accessible communication interface, and a processor-accessible memory. The memory carries instructions executable by the processor to identify user-associated activities. The instructions can also be executed by the processor to transmit at least one command to an optical output device using the communication interface, thereby modulating light from the optical output device based on the identified activity. Summary of the Invention

[0005] The inventors have recognized that lighting systems can transition from one light setting to different light settings, and users can provide feedback related to this change. However, it is possible that the user is dissatisfied with the transition from the first light setting to the second, rather than with the second light setting itself. The learning system can interpret this dissatisfaction as negative feedback on the second light setting, while the user may actually like the second light setting but is unhappy with the way the transition has occurred.

[0006] Therefore, the object of the present invention is to determine whether the feedback provided by the user is for the illumination transition from the first light setting to the second light setting, or whether the feedback is for the second light setting, and thereby improve the learning system.

[0007] According to the first aspect, this objective is achieved by a method for training a machine to automate lighting control actions, wherein the method includes the steps of: controlling one or more lighting devices to present a first lighting effect based on a first set of control parameters; controlling one or more lighting devices to present a second lighting effect based on a second set of control parameters by transitioning from the first lighting effect to a second lighting effect during a transition time period; receiving feedback from a user during a feedback time period; wherein if feedback has been received within a predetermined time period during the feedback time period, the feedback is associated with the transition; and wherein if feedback has been received after the predetermined time period during the feedback time period, the feedback is associated with the second lighting effect; and wherein the method further includes training the machine based on the associated feedback.

[0008] This method may include controlling one or more lighting devices to present first and second lighting effects based on first and second sets of control parameters, respectively. For example, the lighting effect may include the color, color temperature, intensity, beam width, beam direction, illumination intensity, and / or other parameters of the light source from the one or more lighting devices. Control parameters—such as input current, voltage, the orientation of the one or more lighting devices, etc.—may be related to controlling the lighting effect. The one or more lighting devices are arranged to illuminate an environment. A user present in the environment and who has seen the change from the first lighting effect to the second lighting effect after a transition can provide his / her feedback during a feedback period. The method also includes that if feedback has been received within a predetermined period, the feedback may be associated with the transition, and if feedback has been received after the predetermined period, the feedback may be associated with the second lighting effect. This method thus provides a time-based approach to determine whether the feedback provided by the user is for an illumination transition from the first lighting setting to the second lighting setting, or for the second lighting setting itself. This association may be based on time instances in which feedback has been received relative to the transition period. Such assigned feedback can be used to train machines or models, thereby improving the training of machine learning systems. In one example, machine training may include determining a second set of control parameters and / or the transition.

[0009] According to the second aspect, this objective is achieved by a method for training a machine to automate lighting control actions, wherein the method includes the following steps: controlling one or more lighting devices to present a first lighting effect based on a first set of control parameters; controlling one or more lighting devices to present a second lighting effect based on a second set of control parameters by transitioning from the first lighting effect to a second lighting effect during a transition time period; receiving feedback from a user during a feedback time period; assigning probabilities to the association of feedback with the transition and the second lighting effect; wherein during the feedback time period, the probability of the feedback being associated with the transition decreases as a function of time; and wherein the probability of the feedback being associated with the second lighting effect increases accordingly; and training the machine based on the probabilities of the associated feedback.

[0010] In some cases, feedback may be associated with both the transition and the second lighting effect. In such cases (but not limited to these), the method may include assigning probabilities to the association between feedback and the transition and the second lighting effect. Probabilities may include probabilities or weights assigned to the feedback for the transition / second lighting effect. Probabilities may include relative weights or relative probabilities. These probabilities may be based on a time-dependent function, such that the probability of feedback being associated with the transition decreases as a function of time during the feedback period; and the probability of feedback being associated with the second lighting effect increases accordingly. The probabilities (probabilities or weights) of the transition and the second lighting effect can be used separately to train the machine, thereby improving the training of the machine learning system. In one example, a single machine / model can be trained. Additionally and / or alternatively, separate models / machines can be trained for the transition and the second lighting effect.

[0011] In one embodiment, the feedback time period may partially overlap with the transition time period, and the method may further include associating the feedback with the transition if feedback has been received within the overlapping time period.

[0012] The feedback period includes the time during which a user can provide feedback. In other words, user feedback is only acceptable during the feedback period. In this example, the feedback period may partially overlap with the transition period. The feedback period may begin during the transition period and end after the transition period ends. For example, user feedback may be received during the transition period. If feedback has been received during the overlapping period (i.e., during the transition period), it indicates that the user likes / dislikes the transition because the user has not yet seen the second light effect and has already provided his / her feedback. Therefore, this feedback is advantageously associated with the transition. If feedback has been received after the overlapping period (e.g., after the transition period), it can indicate that the user likes / dislikes the second light effect because the user has already seen the second light effect. Therefore, this feedback can be advantageously associated with the second light effect.

[0013] In one embodiment, the feedback period may begin after the transition period. Alternatively, the feedback periods may be non-overlapping and may begin after the transition has occurred, with user feedback only acceptable after the transition has taken place. The timing of user feedback after the transition can determine whether the feedback is associated with the transition or a second lighting effect. The timing of user feedback after the transition can determine the likelihood that the feedback is associated with the transition / second lighting effect.

[0014] In one embodiment, the feedback associated with the transition may be related to transition features, which may include the duration, speed, and / or color of the transition.

[0015] The transition from the first lighting effect to the second lighting effect can be characterized by transition features. These transition features may include the duration, speed, and / or color of the transition. The machine / model can be configured to learn these transition features based on user feedback. This list of features is not exhaustive and does not exclude other transition features.

[0016] In one embodiment, the method may further include determining a second set of control parameters and / or transitions based on previous feedback.

[0017] Machine training can be an iterative process; for example, a second set of control parameters and / or transitions can be based on previous feedback. For instance, if a user prefers a lighting effect and has indicated this via previous feedback, a second set of control parameters can be determined based on those previous effects. Similarly, if the user has indicated a preference for a high brightness level, a second set of control parameters can be determined to keep the brightness level within a high range. Machine training can be iterative based on each piece of feedback.

[0018] In one embodiment, the method may further include determining a second set of control parameters and / or transitions based on predetermined selection criteria.

[0019] The determination of the second set of control parameters and / or transitions can be based on predetermined selection criteria. These criteria can be related to contextual parameters, which may be related to environmental conditions, user, location, and / or time. Contextual parameters can be obtained from (remote) memory and / or detected by one or more sensors. Examples of selection criteria include time of day, date, day of the week, weather conditions, ambient light measurements, occupancy measurements, user activity, and control inputs.

[0020] In one embodiment, the method may further include determining the user's identity; and determining a second set of control parameters and / or transitions based on the determined identity.

[0021] In multi-user environments, where multiple users exist, identifying users and training the machine based on their preferences can be important. In this example, the second set of control parameters and / or transitions (features) can be based on the identified users.

[0022] In one embodiment, feedback may include active feedback or passive feedback.

[0023] User feedback can include proactive or forced feedback that requires the user to provide it. Alternatively, feedback can be passive or non-coercive, which does not require the user to provide feedback, but is learned from the user's behavior.

[0024] In one embodiment, feedback may include active feedback, and active feedback may include user-activated at least one actuator and / or voice input.

[0025] One of the active responses associated with the second set of control parameters and / or transitions may include actuating at least one actuator, such as a like or dislike button. For example, if the user actuates the like button, it is considered positive feedback, and if the user actuates the dislike button, it is considered negative feedback. Additionally and / or alternatively, feedback may also be in the form of a voice command.

[0026] In one embodiment, feedback may include passive feedback, and passive feedback may include feedback based on the user's gaze and / or posture.

[0027] For passive feedback, where no "active" action is expected from the user, the feedback may include the user's gaze and / or posture. In another example, the user's inaction while in the environment can also be considered positive feedback. In an advanced embodiment, the user's EEG may be recorded, and feedback may be based on such a measurement.

[0028] In one embodiment, a machine learning algorithm can be used to train a machine.

[0029] Machine learning algorithms such as supervised learning and / or reinforcement learning can be used to train machines to optimize lighting effects and / or transitions.

[0030] According to the third aspect, the objective is achieved by a controller for training a machine to automate lighting control actions; wherein the controller includes a processor arranged to perform the steps according to the method of the first and / or second aspect.

[0031] According to the fourth aspect, the objective is achieved by a lighting system for training machines to automate lighting control actions, the lighting system comprising one or more lighting devices arranged to illuminate the environment; and a controller according to the third aspect.

[0032] According to the fifth aspect, the objective is achieved by a computer program product including instructions that, when executed by a computer, cause the computer to perform the steps of the methods of the first and / or second aspects.

[0033] It should be understood that computer program products, controllers, and systems may have similar and / or the same embodiments and advantages as the methods described above. Attached Figure Description

[0034] Referring to the accompanying drawings, the above and additional objects, features, and advantages of the disclosed systems, devices, and methods will be better understood through the following illustrative and non-limiting detailed description of embodiments of the systems, devices, and methods, in which:

[0035] Figure 1 An embodiment of a system for training machines to automate lighting control actions is illustrated schematically and exemplary;

[0036] Figure 2 An embodiment of a controller for training a machine to automate lighting control actions is illustrated schematically and exemplary;

[0037] Figure 3 A flowchart illustrating an embodiment of a method for training a machine to automate lighting control actions is shown schematically and exemplaryly;

[0038] Figure 4 A timing diagram for receiving and distributing feedback is illustrated schematically and exemplary;

[0039] Figure 5 A flowchart illustrating, and demonstrating by way of example, is shown as an embodiment of another method for training a machine to automate lighting control actions; and

[0040] Figure 6 Another timing diagram for receiving and distributing feedback is shown schematically and exemplary.

[0041] All figures are schematic and not necessarily to scale, and generally only show the parts necessary to illustrate the invention, where other parts may be omitted or merely suggested. Detailed Implementation

[0042] Figure 1An embodiment of a system 100 having one or more lighting devices 110a-110d for illuminating an environment 101 is illustrated schematically and exemplary. Environment 101 can be an indoor or outdoor environment, such as an office, factory, residence, grocery store, hospital, sports field, etc. System 100 exemplaryly includes four lighting devices 110a-110d. Lighting devices 110a-110d may be included in a lighting system. The lighting system may be a connected lighting system, such as Philips Hue, where lighting devices 110a-110d can be connected to an external network, such as the Internet. Lighting devices 110a-110d are devices or structures arranged to emit light suitable for illuminating environment 101, providing or substantially contributing to illumination of a scale sufficient for that purpose. Lighting devices 110a-110d include at least one light source or lamp (not shown), such as an LED-based lamp, gas discharge lamp, or incandescent lamp, etc., and (optionally) have associated supports, housings, or other such enclosures. Each of the lighting fixtures 110a-110d can take any of various forms, such as ceiling-mounted lighting fixtures, wall-mounted lighting fixtures, wall washer lights, or freestanding lighting fixtures (and the lighting fixtures do not necessarily have to be the same type). In this exemplary figure, lighting fixtures 110a-110c are ceiling-mounted, and lighting fixture 110d is a freestanding lighting fixture. System 100 can contain any number / type of lighting fixtures 110a-110d.

[0043] Lighting devices 110a-110d can be controlled based on a first set of control parameters. Control of lighting devices 110a-110d may include controlling one or more of the following: the color, color temperature, intensity, beam width, beam direction, illumination intensity, and other parameters of one or more light sources (not shown) of lighting devices 110a-110d. Lighting devices 110a-110d can be controlled based on a second set of control parameters. When lighting devices 110a-110d are controlled based on the first and second sets of control parameters respectively, first and second lighting effects can be presented. The second set of control parameters may differ from the first set of control parameters, such that the difference between the first and second lighting effects can be perceived by user 120. In a simple example, the lighting effect is the brightness level of lighting devices 110a-110d; for example, the first lighting effect is a brightness level of 30%, and the second lighting effect is a brightness level of 70%. The second lighting effect (i.e., the 70% brightness level) is determined such that the difference between the first and second lighting effects can be perceived by user 120. For example, the selection of a 70% brightness level is based on the ambient light level in environment 101, allowing user 120 to perceive a 50% difference in brightness level. In another example, lighting devices 110a-110d are controlled not to provide light output based on a first set of control parameters.

[0044] For the change from the first lighting effect to the second lighting effect, there is a transition phase in the transition time period. For example, the first lighting effect at a brightness level of 30% transitions slowly to the second lighting effect at a brightness level of 70% by linearly increasing the brightness level. In another example, the transition can include an exponential change. In an extreme example, the transition time period is zero, so that the first lighting effect at a brightness level of 30% is changed to the second lighting effect at a brightness level of 70% in the next instant.

[0045] In one example, lighting effects include lighting scenes that can be used to enhance entertainment experiences such as audiovisual media, setting the mood and / or emotion for the user. For example, for a Philips Hue-connected lighting system, the first lighting effect is a 'Magic Forest' lighting scene, and the second lighting effect is a sleep-inducing lighting scene. The first and / or second lighting effects can include static lighting scenes. The first and / or second lighting effects can include dynamic lighting scenes, where dynamic lighting scenes include lighting effects that change over time. For dynamic lighting scenes, the first and / or second lighting effects can include a first lighting state and a second lighting state. The first lighting state can include a first (predefined) pattern, and the second lighting state can include a second (predefined) pattern. The pattern can include the duration of the lighting effect, the level of dynamics, etc. The first and second sets of lighting states can be associated with first and second subsets of a second set of control parameters, respectively. In such an example, machine training includes automating the (first and / or second) subsets of the second set of control parameters. For dynamic lighting effects, transitions can include, for example, changing color, dynamics, presenting intermediate scenes, etc.

[0046] Feedback is received from the user during the feedback period. Feedback can be active or passive. Active feedback can include actuating at least one actuator, such as: a like / dislike button on the user's mobile device 136 indicating his / her preference; a wall switch 130, which, for example, controls lighting devices 110a-110d to change a second lighting effect to (implicitly) another lighting effect indicating the user's preference (dislike) for the second lighting effect; and / or via the user's voice input 133. The at least one actuator can be used to control lighting devices 110a-110d.

[0047] Feedback may include passive feedback, and the passive feedback includes feedback based on the user 120's gaze and / or posture. System 100 may include sensing devices 140, such as presence sensors, gaze detection devices, such as those using RF sensing. Methods for detecting gaze and / or posture are well known in the art and therefore will not be discussed further here.

[0048] Feedback from user 120 can be received during a feedback period. The feedback period can begin after a transition has occurred (e.g., after a transition period) or partially overlap with a transition period, such that user 120's feedback is acceptable and not discarded during the transition. Based on this condition, if feedback has been received within or after a predetermined period during the feedback period, the feedback is associated with a transition or a second light effect. Alternatively, probabilities are assigned to the association of feedback with a transition and a second light effect; wherein the probability of feedback being associated with a transition decreases as a function of time during the feedback period; and wherein the probability of feedback being associated with a second light effect increases accordingly. For example, when the feedback period partially overlaps with a transition period, and if feedback has already been received within the overlapping period, the feedback is assigned to a transition.

[0049] The machine can be trained based on associated feedback. Machine learning algorithms—such as supervised learning (e.g., SVM, decision forest, etc.)—can be used to train the machine. Reinforcement learning can be used to train the machine. The learning algorithm can include iterative learning, such that the determination of a second set of control parameters and / or transitions can be based on previous feedback, where the algorithm iteratively trains the machine. Training can include different phases, such as a feedback phase, in which feedback is received from user 120. The length of the feedback phase can include a feedback period, which is assumed to be long enough to capture sufficient information for training. After the feedback phase, a training phase can begin. A training phase can be defined, during which the machine is trained. In one example, two different machines can be trained for the second lighting effect and the transition. For iterative learning, the feedback phase and the training phase can be used iteratively. In one example, the feedback phase could be the first week after lighting devices 110a-110b have been initially installed, networked, and user 120 has begun using them. The duration of the feedback phase can be defined by user 120. In one example, the training phase may include a learning phase and a fine-tuning phase; wherein, in the learning phase, a second set of control parameters and / or transitions may be learned based on user feedback. In the fine-tuning phase, the second set of control parameters and / or transitions may be further optimized based on additional user input. In one example, the identity of user 120 is determined, for example, by an imaging sensor. The determination of the second set of control parameters and / or transitions may be based on the determined identity, such as based on the preferences of the identified user. In one example, the second set of control parameters and / or transitions may be based on predetermined selection criteria. Selection criteria may include, for example, time of day, date, day of the week, weather conditions, ambient light measurement, occupancy measurement, user activity, and control inputs, etc.

[0050] Figure 2An embodiment of a controller 210 for training a machine to automate lighting control actions is illustrated schematically and exemplary. The controller 210 may include an input unit 214 and an output unit 215. The input unit 214 and output unit 215 may be included in a transceiver (not shown) arranged to receive (input unit 214) and transmit (output unit 215) communication signals. The communication signals may include control commands for controlling lighting devices 110a-110d. The input unit 214 may be arranged to receive communication signals from a switch 130 and / or from a voice command 133. The input unit 214 may also be arranged to receive communication signals from a user mobile device 136. The communication signals may include control signals. The controller 210 may further include a memory 212, which may be arranged to store communication IDs of lighting devices 110a-110d and / or sensors 140, etc. The memory 212 may also be arranged to store previous feedback. The controller 210 may include a processor 213 for training the machine. Machine training may be performed externally to the controller 210. Controller 210 can be used to infer user preferences based on a trained machine. In one example, the inference can also be performed outside of controller 210, and controller 210 can be configured to receive control commands based on the trained machine.

[0051] Controller 210 can be implemented in a separate unit from lighting fixtures 110a-110d, sensor 140, and wall switch 130, such as a wall panel, desktop computer terminal, or even a portable terminal (e.g., laptop, tablet, or smartphone). Alternatively, controller 210 can be incorporated into the same unit as sensor 140 and / or the same unit as one of lighting fixtures 110a-110d. Furthermore, controller 210 can be implemented in or away from environment 101 (e.g., on a server); and controller 210 can be implemented in a single unit or as a distributed function distributed across multiple independent units (e.g., a distributed server comprising multiple server units in one or more geographical locations, or a distributed control function distributed between lighting fixtures 110a-110d or between lighting fixtures 110a-110d and sensor 140). Furthermore, the controller 210 may be implemented as software stored in memory (including one or more memory devices) and arranged to execute on a processor (including one or more processing units), or the controller 210 may be implemented as dedicated hardware circuitry, or configurable or reconfigurable circuitry such as a PGA or FPGA, or any combination thereof.

[0052] Regarding the various communications involved in achieving the above functions, for example, in order for the controller 210 to receive the presence signal output from the presence sensor 140 and control the light output of the lighting devices 110a-110d, these can be achieved by any suitable wired and / or wireless means: for example, by means of a wired network, such as Ethernet, DMX network or Internet; or a wireless network, such as a local (short-range) RF network, such as Wi-Fi, ZigBee or Bluetooth network; or any combination of these and / or other means.

[0053] Figure 3 A flowchart illustrating an embodiment of a method 300 for training a machine to automate lighting control actions is shown schematically and exemplary. Method 300 may include controlling one or more lighting devices 110a-110d (310) to produce a first lighting effect based on a first set of control parameters. Method 300 may also include controlling one or more lighting devices 110a-110d (320) to produce a second lighting effect based on a second set of control parameters by transitioning from the first lighting effect to a second lighting effect during a transition time period. The transition may be characterized by transition features such as duration, speed, color, etc.

[0054] Method 300 may further include receiving 330 feedbacks from the user during the feedback period. The feedback may be active or passive. In one example, a signal indicating the user's presence may be received, and then a second light effect with a transition from the first light to the second light effect may have been presented. The feedback period may include sufficient time for the user to observe / perceive the transition and the second light effect, as well as sufficient feedback to train the model.

[0055] Method 300 may further include the following condition: if the predetermined time period t_t during the feedback time period (e.g. Figure 4 If feedback 330 has been received (yes, condition 343), then the feedback is associated with the transition 350. Now refer to Figure 4 The diagram schematically and exemplaryly illustrates a timing diagram for receiving and distributing feedback. Figure 4The x-axis represents the time axis t, and the y-axis represents the presentation and change of the light effect after the transition. During the time period t0-t1, one or more lighting devices 110a-110d are controlled based on a first set of control parameters to present a first light effect. Depending on the light effect, the time period t0-t1 can be seconds, minutes, or hours (or even longer). By transitioning from the first light effect to a second light effect during the transition time period t1-t2, one or more lighting devices 110a-110d are controlled during the time period t2-t3 to present a second light effect based on a second set of control parameters. The duration of the time period t2-t3 can depend on the second light effect. The transition time period t1-t2 can depend on the transition presented. In one example, the transition time period t1-t2 can be a feature learned based on the preferences of user 120. In one example, the transition time period t1-t2 can be close to zero, such that one or more lighting devices 110a-110d are controlled 320 to transition instantaneously to the second light effect. The association between feedback and transition or second light effect can be based on a predetermined time period t_t. Therefore, the determination of whether feedback is associated with transition or second light effect can be based on the time instance (e.g., t_t) at which feedback is received from user 120.

[0056] Looking back now Figure 3 As previously described, the method may include: during the feedback time period, if feedback has been received within a predetermined time period t_t (yes, condition 343), then the feedback is associated with a transition (350). Alternatively, if feedback 330 has been received after a predetermined time period during the feedback time period, then the feedback is associated with a second lighting effect (360). The predetermined time period may be determined based on, for example, user history data, the duration of the second lighting effect, the duration of the transition, etc. The predetermined time period may be randomly selected.

[0057] Method 300 can further include training the 370 machine based on the associated 350-360 feedback. Machine learning algorithms can be used to train the machine. For example, supervised learning can be used. Supervised learning is a machine learning task that learns a function or model that maps inputs to outputs based on input-output data pairs. It infers a function from a labeled training dataset that includes a set of training data. In supervised learning, each sample in the training dataset is a pair consisting of an input (e.g., a vector) and a desired output value. For example, the output is associated with 350-360 feedback, and a second set of control parameters and / or transitions is the input vector. The training dataset includes the output (feedback) and the input (second set of control parameters / transitions). Supervised learning algorithms, such as Support Vector Machines (SVM), Decision Trees (Random Forests), etc., are used. The training dataset is analyzed and an inferred function or model is generated, which can be used to make predictions based on new datasets. In this example, a binary classifier machine can be trained to predict user 120's preferences for a new set of control parameters and / or transitions. A single model can be trained for both the second set of control parameters and the transition; alternatively, two separate models can be trained for each of the second set of control parameters and the transition. If the model predicts a positive user preference for the new set of control parameters, the lighting devices 110a-110d can be controlled based on the new set of control parameters to produce new lighting effects. As an alternative to supervised learning, reinforcement learning can be used to train the machine. Other learning algorithms—such as rule-based learning, probabilistic reasoning, and fuzzy logic—can also be considered to train machines known in the art for automating lighting control actions.

[0058] In different examples, if it cannot be determined whether the feedback is related to the transition or the second light setting, the user can be asked to clarify this (e.g., via a voice assistant, via the display of the mobile device 136, etc.).

[0059] Figure 5 A flowchart illustrating, and demonstrating, is shown schematically and exemplary of an embodiment of another method 500 for training a machine to automate lighting control actions. The method steps of controlling 310-320 and receiving feedback from user 120 are similar to... Figure 3 The same method as mentioned in method 300. Method 500 may further include assigning a probability 540 to the association of feedback with transition and second light effect; wherein the probability 540 of feedback being associated with transition decreases as a function of time during the feedback time period; and wherein the probability 540 of feedback being associated with second light effect increases accordingly. The probability of feedback may include probability values ​​that can be assigned 540 to transition and second light effect. For example, since the sum of probabilities is 1, depending on when the event has received feedback, a probability of 0.7 can be assigned to transition and a probability of 0.3 can be assigned to second light effect. Now refer to Figure 6 The diagram schematically and exemplaryly illustrates the timing diagram used for receiving and distributing feedback. The x and y axes, along with the time periods t0-t1 (first lighting effect), t1-t2 (transition period), and t2-t3 (second lighting effect), are shown in conjunction with... Figure 4 The same applies as shown in the diagram. The increase / decrease in probability is exemplarily illustrated by dashed lines 610 and 620. In this exemplary diagram, the probability of user feedback associated with the transition decreases linearly over time 620. Similarly, in this exemplary diagram, the probability of user feedback associated with the second lighting effect increases linearly over time 610. In another example, the probability of feedback on the transition / second lighting effect may increase / decrease exponentially or with a different function. In one example, the increase and decrease may include different time functions, such as linear increase and exponential decrease. At a given point in time, the probability for both the transition and the second lighting effect may be the same (50%-50%).

[0060] Looking back now Figure 5 Method 500 can further include training 550 machines based on the probabilities of associated feedback. Training 550 takes probabilities into account during the training process. The difference from the algorithms mentioned earlier (e.g., supervised learning, reinforcement learning, etc.) lies in the fact that each feedback may not be given equal weight, but rather adjusted based on probability. These probability values ​​can be used separately for each transition and second light effect of each model, or they can be used as relative probabilities during training 550.

[0061] When the computer program product is run on the processing unit of a computing device (such as controller 210), methods 300-500 can be executed by the computer program code of the computer program product.

[0062] It should be noted that the above embodiments are illustrative and not limiting of the invention, and those skilled in the art will be able to devise many alternative embodiments without departing from the scope of the appended claims.

[0063] In the claims, any reference numerals placed between parentheses should not be construed as limiting the claims. The use of the verb "comprising" and its variations does not exclude the presence of elements or steps other than those stated in the claims. The article "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer or processing unit. In an apparatus claim enumerating several means, several of these means may be embodied by the same hardware item. The mere fact that certain measures are referenced in mutually different dependent claims does not indicate that a combination of these measures cannot be used advantageously.

[0064] Various aspects of this invention can be implemented in a computer program product, which may be a collection of computer program instructions stored on a computer-readable storage device and executable by a computer. The instructions of this invention can be any interpretable or executable code mechanism, including but not limited to scripts, interpretable programs, dynamic link libraries (DLLs), or Java classes. The instructions may be provided as a complete executable program, a partial executable program, as a modification (e.g., an update) of an existing program, or as an extension (e.g., a plugin) of an existing program. Furthermore, some processing of this invention may be distributed across multiple computers or processors or even the “cloud.”

[0065] Storage media suitable for storing computer program instructions include all forms of non-volatile memory, including but not limited to EPROM, EEPROM, and flash memory devices, disks such as internal and external hard drives, removable disks, and CD-ROMs. Computer program products may be distributed on such storage media or made available for download via HTTP, FTP, email, or through a server connected to a network such as the Internet.

Claims

1. A method for training a machine to automate lighting control actions, wherein the method includes the following steps: - Control one or more lighting devices based on the first set of control parameters to present a first lighting effect; - By transitioning from the first lighting effect to the second lighting effect during a transition period, the one or more lighting devices are controlled to present the second lighting effect based on a second set of control parameters; - Receive feedback from users during the feedback period; Its features Wherein, if the feedback has been received within a predetermined time period during the feedback time period; - Associate the feedback with the transition; and Where the feedback has been received after a predetermined period of time during the feedback period; - Associate the feedback with the second light effect; and The method further includes: - The machine is trained based on associated feedback; wherein the transition is characterized by transition features, and wherein machine training includes determining the second set of control parameters and / or the transition features.

2. The method of claim 1, wherein the feedback time period partially overlaps with the transition time period, and wherein the method further comprises: If the feedback has been received within the overlapping time period; - Associate the feedback with the transition.

3. The method of claim 1, wherein the feedback time period begins after the transition time period.

4. The method of claim 1 or 2, wherein the feedback associated with the transition is related to transition features, wherein the transition features include the duration, speed and / or color of the transition.

5. The method according to claim 1 or 2, wherein the method further comprises: - Determine the second set of control parameters and / or the transition based on previous feedback.

6. The method according to claim 1 or 2, wherein the method further comprises: - Determine the second set of control parameters and / or the transition based on predetermined selection criteria.

7. The method according to claim 1 or 2, wherein the method further comprises: - Determine the identity of the user; and - Determine the second set of control parameters and / or the transition based on the identified identity.

8. The method according to claim 1 or 2, wherein the feedback includes active feedback or passive feedback.

9. The method of claim 1 or 2, wherein the feedback includes active feedback, and wherein the active feedback includes at least one actuator actuated by the user and / or voice input.

10. The method of claim 1 or 2, wherein the feedback includes passive feedback, and wherein the passive feedback includes feedback based on the user's gaze and / or posture.

11. The method of claim 1 or 2, wherein a machine learning algorithm is used to train the machine.

12. A method for training a machine to automate lighting control actions, wherein the method includes the following steps: - Control one or more lighting devices based on the first set of control parameters to present a first lighting effect; - By transitioning from the first lighting effect to the second lighting effect during a transition period, the one or more lighting devices are controlled to present the second lighting effect based on a second set of control parameters; - Receive feedback from users during the feedback period; Its features - Assigning probabilities to the association of the feedback with the transition and the second light effect; wherein during the feedback time period, the probability of the feedback being associated with the transition decreases as a function of time; and wherein the probability of the feedback being associated with the second light effect increases accordingly; and - The machine is trained based on the probability of associated feedback; wherein the transition is characterized by transition features, and wherein machine training includes determining the second set of control parameters and / or the transition features.

13. The method of claim 12, wherein the feedback time period begins after the transition time period.

14. The method of claim 12, wherein the feedback associated with the transition is related to transition features, wherein the transition features include the duration, speed, and / or color of the transition.

15. The method of claim 12, wherein the method further comprises: - Determine the second set of control parameters and / or the transition based on previous feedback.

16. The method of claim 12, wherein the method further comprises: - Determine the second set of control parameters and / or the transition based on predetermined selection criteria.

17. The method of claim 12, wherein the method further comprises: - Determine the identity of the user; and - Determine the second set of control parameters and / or the transition based on the identified identity.

18. The method of claim 12, wherein the feedback includes active feedback or passive feedback.

19. The method of claim 12, wherein the feedback includes active feedback, and wherein the active feedback includes at least one actuator actuated by the user and / or voice input.

20. The method of claim 12, wherein the feedback includes passive feedback, and wherein the passive feedback includes feedback based on the user's gaze and / or posture.

21. The method of claim 12, wherein a machine learning algorithm is used to train the machine.

22. A controller for training a machine to automate lighting control actions; wherein the controller includes a processor configured to perform the steps of the method according to any one of claims 1-21.

23. A lighting system for training machines to automate lighting control actions, comprising: - One or more lighting fixtures are arranged to illuminate the environment; - The controller according to claim 22.

24. A computer program product comprising instructions that, when executed by a computer, cause the computer to perform the steps of the method according to any one of claims 1-21.

Images