.. , 71742.607 The present invention relates to a switching element which can be used for controlling a target variable/quantity, for example of an electrically-driven apparatus, and a process for its control. By way of example, there can be specified switching elements for the control of light switches, alarm systems or other household appliances, such as coffee machines or entertainment systems. In addition, the control of washing machines, computers or micro-machines can be advantageously carried out with the switching element according to the invention.
Time switches are known, but these have to be labouriously programmed by the user. It has been discovered that the effort of programming many household devices is so great for the average user that the possibilities of a programmable apparatus are not fully exploited.
A programmable switching unit is known from EP 0,581,555 A1, where the unit can be switched backwards and forwards between a learning mode, in which a table with on or off states of the appliance which is to be controlled is updated over a certain time period, and a second, automatic, mode. By logical combination of the stored switching data, a sequence of switching processes is generated which corresponds to the previous switching behaviour. In this way, it is possible to simulate normal operation of the switch, even if the user is not present. One example of an application of the switching element is the simulation of the switching on and off of lights in a house in order to give the impression from the outside that the house is occupied.
Such a switching element, however, is merely able to simulate or control the switching processes dependent on time inputs. Simulation of a target variable which is to be switched, the value of which depends on outside influences, cannot be realised with such a switch. An example of this is control of an irrigation system.
Such a system is typically operated dependent on the air humidity, the humidity of the ground, brightness or time of day. A switching element as described above would not be in a position to implement suitable control of such an irrigation system.
A switching unit is known from US 5,608,843 which can generate an estimation function for learning purposes.
Finally, DE 295 O1 306 describes.an automatic switching unit, which operates by means of a neuro-fuzzy-logic circuit, which can be operated manually and also automatically, influenced by sensor signals.
The object of the invention is to provide a switching element via which control of a target variable can be implemented, dependent on outside influence variables, in a simple manner by means of switching instructions.
This object is solved by a switching element with the features of claim 1 and the process with the features of claim 6.
Firstly, with the switching element according to the invention it is possible, as a result of external e.g.
manually-entered switching instructions for control of a target variable, to calculate an estimation function for representation of the target variable, dependent on the influence variables of the target variable. Here, a memory unit of the switching element notes the values of the influence variables on entry of a switching instruction during a learning phase. When an adequate statistic basis is compiled, the estimation function can be generated with sufficient exactness for simulation of the previous switching behaviour. After generation of such an estimation function the switching element can produce internal switching instructions to control the target variable. This will be explained again with reference to the above example of an irrigation system.
During the learning phase, and according to the requirements, the user gives instructions for turning the system on or off. These switching instructions take place under certain external conditions, for example a certain air humidity, a certain ground humidity and a certain brightness. The values of these so-called influence variables are determined by sensors and are transferred to the switching element. For each on or off instruction a value tupel comprising the in- or output instruction and the respective values of the influence variables is stored in a storage unit. An estimation function is produced by means of suitable mathematical processes if a sufficient number of such tupels are present. After switching the switching element from the learning phase to an automatic phase, the switching element is then in a position to independently generate on or off instructions. The corresponding sensor values are detected and evaluated at regular intervals during the automatic operating state. The value of the estimation function is detected corresponding to the detected influence variables which represent variables of the estimation function which is dependent thereon. It is possible to e.g. use one of the influence variables as feedback variables.
An estimation function typically takes in values between 0 (off state) and 1 (on state). If the value of the estimation function determined for the current values has a distance from the value 0 or 1 which falls below a certain threshold distance, a switching-on or switching-off instruction is generated. Such distance values, where a switching instruction is generated if they are not met, can be set in any manner. It should be noted that not only discrete but also continual values of a target variable can be controlled. A plurality of target variables, which depend entirely or partially on the same influence variables, can also be controlled by means of a switching element according to the invention.
It is also conceivable to control target variables, the influence variables of which are independent of one another, by means of a switching element.
The switching element, according to the invention, determines, at regular intervals or continually, the influence variables which influence the target values which are to be controlled by it. In the event of user actions, i.e. an external input of a switching instruction, the switching element takes note of the circumstances in which these have taken place, i.e. at which values of the influence variables. On the basis of these inputs, a system attempts to simulate and copy the control behaviour.
Advantageous embodiments of the switching element, according to the invention, and the process for controlling the switching element, according to the invention, are the subject of the subordinate claims.
Advantageously, determination and storage of the values of the at least one influence variable of the target variable take place concurrently,with the input of an external switching instruction. This means, for example, that in the event of manual operation of the switching element, the values of the influence variables present at this time are detected. In this way, an estimation function with the required precision can r easily be calculated if there are a sufficient number of switching operations. It is also conceivable that the switching element automatically determines the current value of the influence variables at a particular point between two external switching instructions, and saves these together with the current switch state. In this way, an adequate statistical basis for the generation of an estimation function can be attained more rapidly.
Advantageously, the switching element has means for automatic or externally-specified switching from a first operating state, in which the switching element can exclusively be controlled by external switching instructions, to a second operating state in which it can be controlled by internal and/or external switching instructions. During the first operating state, i.e.
during a learning phase, the switching element stores all switching instructions entered externally by a user.
It is only on switching to a second operating state, i.e. an automatic phase, that a generated estimation function is used for automatic control of the switching element. This externally-specified changeover can take place by means of a switch. It is also possible to program the switching element in such a manner that it detects itself a time which is suitable for changeover, for example whilst monitoring is being carried out as to whether a sufficient statistical basis is present for the generation of an estimation function.
Advantageously, the switching element, according to the invention, has at least one output via which an output signal, especially a signal corresponding to a completed switching instruction, can be set to an output of a further switching element as an influence variable of the switching element. In this way, it is possible to crosslink the switching elements, according to the invention, with one another in any manner. In doing this, the switching elements can be connected to one another in series, in a tree structure or in rings, i.e.
coupled. Here, an output signal of a switching element is switched to one of the inputs of another switching element. In particular, a switching element can serve as an input signal transmitter for any number of other switching elements. Hence the output signal of the switching element serves as the influence variable of a further switching element.
According to a preferred embodiment the switching element, according to the invention, has means for assessing a threshold value, by which a certain correlation of the estimation function and a switching instruction, which is to be generated internally, can be set. By setting a threshold value of this kind, it is possible to set the switching element "conservatively"
or also "progressively". If the threshold value is set conservatively, an internal switching instruction is only generated when the value of the estimation function for the current influence variable lies very close to the function values 1 (corresponding to switching on) or 0 (corresponding to switching off).
The present invention will now be described in detail, especially with reference to the diagram. Here, the sole figure shows a schematic block representation of an embodiment of the switching element according to the invention.
With reference to the Figure, the switching element according to the invention has a switch 10 for external input of switching instructions. The target variable can be controlled, or switched on and off, by means of this switch 10 via a load switch element 11. The load switch element connects, or interrupts, a current signal input St-E with regard to a current signal output St-A1.
Hence this signal St-A1 corresponds to the target variable which has to be controlled.
Furthermore, the switching element has inputs 12 via which influence variables E1 to En of the target variable detected by sensors, can be read in. Inputs E1 to En can in each case be provided with AD converters AD-1 to AD-n, as in the represented embodiment example.
Such AD converters are naturally not necessary if influence variables are entered which are already digitised.
Furthermore, the switching element has a calculation and storage unit 13 (surrounded by a dotted line). This comprises a microprocessor ~.P, a ROM memory, a RAM
memory and a timer. These elements are connected together by means of a ~,P bus. In each case, the switching instructions are stored in the RAM memory with the values of the influence variables (n-tupel) which correspond to them. The ROM memory contains the programs for generating an estimation function from the values stored in this manner, and also the programs for controlling the switching element.
Furthermore, the switching element embraces a second output 15 via which a signal St-A2 can be transmitted, for example to a second switching element. The output signal St-A2 can here serve as an influence variable of a second switching element. The output 15 can be provided with a DA converter DA-1.
Furthermore, the switching element is provided with an auto-key 17 via which a changeover is possible from a first operating state, or a learning phase, into a second operating state, or an automatic phase.
The function of the switch, according to the invention, .c, _ g _ will be described in the following text.
Firstly, manual switching instructions are entered via the switch 10 during the learning phase. The values of the influence variables present at inputs E1 to En when the switching instructions are given are also determined here. Data tupels, comprising the respective switching instruction and the corresponding values of the influence variables, are stored in the RAM. If a sufficient number of these data tupels is present, an estimation function is generated for representation of the switching instruction or the target variable which is to be controlled by this, dependent on the stored influence variables. Such a switching function will typically oscillate between function values 0 and 1.
Here, the function value 1 denotes a "secure" switching on of the switching element for a certain influence variable correlation, and correspondingly a function value 0 denotes a "secure" switching off function value.
Naturally, the estimation function can also oscillate between any other values. Therefore it can be assumed, if such influence values are present, that the user of the switching element would have securely made an on or off switching instruction. Mathematical processes for generating such an estimation function are described below.
If the system is now switched to the second operating state, i.e. the auto-phase, by operation of the switch 17, the switching element carries out an automatic control of the target variable St-A1. To this end, the influence variables E1 to En are detected continually or at discrete intervals. The corresponding value of the estimation function is determined for the influence variables present in each case, which represent the n-variables of the estimation function. If this _ g _ estimation function value lies close enough to the value 1 or 0, a corresponding on or off switching instruction is created by the microprocessor and given to the load switch element. The correlation or required proximity of the estimation function value to the function values 1 or 0 can be set by means of the threshold value switch 16. It is possible to regulate the switching characteristics individually, namely conservatively or progressively, by pre-setting the threshold value. The threshold value here defines a line between the function values 0 and 1. Complex estimation functions or threshold value functions are also conceivable where, for example a plurality of function values are detected and correlated together.
In the following text, mathematical processes will be given by means of which estimation functions can be created. Each mathematical process must here generate an estimation function from a quantity of random samples or measurements which comprise the influence variables or a target variable.
The state at the point in time of the switching instruction is described by the quantity of the influence variables which can accept either continual or discrete values. External parameters which can be entered into the system via the inputs E1 to En are, for example, temperature, air humidity, brightness, entry of an event or input of an output signal from a further switching element.
Furthermore, the system can generate internal, timer-controlled parameters, for example Fourier base functions (trigonometrical functions). Here, the time is measured relatively, i.e. t = 0 is any point in time.
The advantage of this is that no clock must be set from outside. The various influence variables are henceforth designated with ei, i = l...n. A target variable b is assigned to a set, or a tupel, of influence variables ei.
It is assumed that j random samples, or switching instructions, are stored in the system. Each random sample is therefore comprised of a target parameter value b and n influence variables ei.
Several possibilities exist for calculation of an estimation function on this basis. For example, it is possible to generate a linear combination of the state parameter of the form S: - E xiei, i = 1, . . .,n Here, xi represents the weighting of the linear estimation function.
However, it is also possible to generate linear combinations of non-linear transformations of the state parameters or non-linear combinations of non-linear transformations of the state parameters.
Common to the processes is that it is estimated on the basis of an optimisation criterion, i.e. that the state parameters should be combined in such a way that the target variable is estimated as well as possible. The least square method (linear regression analysis), for example, can be used for this. A coefficient xi must be selected in such a way that the sum Sj-b~ ~~ . 7 - 1. . . . ,m is as low as possible. Here, the double strokes represent the Euclidic norm.
However, spline interpolation processes and processes of neuronal networks, or also fuzzy logic methods or variance-reducing processes, are also conceivable.
Suitable processes are described in the applicant's older patent application 19600694.5. These processes are expressly also made subject of the present 5 application.
In the following text, preferred fields of application for the switching element, according to the invention, will be specified.
Generally, the switching elements, according to the invention, can be used wherever time switches are to be replaced, for example in an intelligent or a learning manner. Here, pre-setting need not be carried out by an 15 installation engineer or a user.
The use of switching elements proves advantageous where processes must be controlled in a complex manner i.e.
where two or more predominantly-independent input 20 parameters determine a target variable (for example temperature, brightness or air humidity in the case of a garden irrigation system) and contradict each other, or are strongly weighted in a different manner, with regard to their control.
Furthermore, the use of the switching elements proves useful where control is not immediately transparent to a user because e.g. not all influences have been considered, e.g. when establishing time-controlled 30 services in operating systems. The switching element, according to the invention, can also advantageously be used to balance interests, e.g. of several staff members in an office. An example of this is roller blind control in an office with several staff members who have 35 different opinions with regard to heat or sunlight conditions.
It is also possible, especially in the case of crosslinking a number of switching elements according to the invention, to activate or control processes with complex subsidiary conditions. An example of this would be intelligent or crosslinked building units in the field of building insulation technology.
Preferred applications of the switching elements, according to the invention, are the control of coffee machines and also the control of televisions and video recorders. Here, conventional time clocks can be replaced. In the case of the control of TVs and video recorders, especially with identification of certain programmes, a pre-selection or filtering process can be carried out.
In addition, the control of washing machines, answerphones or telephone systems is simplified by the switching elements according to the invention.
The switching elements, according to the invention, can be used in the computing sector. Switching elements designed as software modules are especially suitable.
For example, every time-controlled operating system/application programme service can intelligently be replaced by a switching element (software time switch). Data save functions or updates of user profiles can be controlled. Furthermore, for example, it is possible to start an end-of-day procedure only on certain days. Usually, the setting up of an operating system environment with various services is very labourious. A learning system which uses the time switch, according to the invention, automatically recognises exception rules, special features or simple routine operations, for example relating to weekdays, without these having to be concretely entered by the user.
For example, a switching element can sit in the server in a client-server environment and can be assigned to a tape machine. It learns when, and under which conditions, the user activates the tape back-up and then later carries out these tasks automatically. After a learning phase, other operating system processes can be triggered by a cascaded interconnection of switching elements on an operating system level, taking into consideration further subsidiary conditions, wherein this is realised by means of further inputs in subordinated switching elements.
As mentioned as an example in the introduction, the switching elements, according to the invention, can be used in field of irrigation technology. Here, especially external temperature, brightness and air humidity can be named as external input variables. Time can be used as an internal input variable. It proves advantageous for control of the system to further provide, for example, a feedback input variable, for example a humidity sensor which measures the humidity of the ground.
The switching elements can also be used to control a hot water storage tank. The switching element continually creates a use profile by means of the hot water "in demand" without the necessity of being set up by a user and, after the learning phase, keeps a corresponding amount of hot water in reserve. According to the behaviour learnt from the user, for example, the storage tank keeps a supply of hot water corresponding to its capacity which is reinforced during the day and is generated to a lesser degree, or not at all, during the night. In this way, substantial savings on energy are possible. The external input variable here is the demand for water. The internal input variable is again the time. Water temperature and water level can .i CA 02298460 1999-12-23 advantageously be used as the feedback input variable.
Here, the target variable is the decision as to whether, and how much, hot water should be produced. A time-dependent use profile is produced. Pre-setting of such a use profile on site is conceivable. A water boiler could also be equipped with the switching element, according to the invention. For example, it is conceivable that the water throughput (which can be monitored by means of a sensor), the attained temperature and the internal time are here used as influence variables. Here, the target variable is a decision as to whether the water should be heated or not.
In addition, it is conceivable to use the switching elements, according to the invention, in micro-machines or micro-pumps. It is generally not easy to set micro-machines from outside, since very large switches are required compared to the machine itself. It is conceivable that switching elements, according to the invention, are used where the learning or automatic phase can be set from outside by different light, special acoustic signals, etc.
The switch, according to the invention, can be used in the field of robotics, for example in the toys sector.
Up to now, it was common for system toy components (i.e.
individual building blocks of assemblable systems) to firstly be programmed at the computer, for example in the sense of an explicit specification of behaviour.
Components programmed in this way are used in toy configurations and control the toy behaviour or environment on the basis of attached sensors (for example, a sufficiently-intense signal in a radar sensor provided in a toy car, caused by proximity to an object causes the car to come to a halt).
_i, CA 02298460 1999-12-23 It is possible, according to the invention, to make the toy environment from switching elements. For example, a switching element is assigned to each leg of a robot animal which controls a movement of the respective leg.
Additionally-provided sensors (distance, light, temperature sensors or cameras) can be used as influence variables of the elements provided as switching elements. The internal time and the outputs of the remaining switching elements are also conceivable as further influence variables. The target variable or target function, e.g. of a switching element, is, for example, the control of a motor or servo motor controlling the leg.
For example, it is conceivable that the toy animal moves in an arbitrary manner during a first learning phase. A
user can specify via a control switch assigned to the animal as to whether the current behaviour is adequate or not. For example, the animal moves towards an obstacle. It is specified, by operation of the control switch (at a certain distance of the animal from the obstacle) that a further approach towards the obstacle is not adequate. This can apply in a similar manner to the switching on of a light. It can be specified, by operating the control switch, that in this case movement towards the light is to take place. It is possible to give weight to the respective influences by the individual switching elements being controlled on the basis of the environment influence variables and the output signals of the remaining switching elements.
Over time, the toy animal learns to "behave" in a manner desired by the user.
Television-child control also proves to be an advantageous field of application for the switching elements. With a switching element used in a television for authorising only those programmes which are desirable for children, only those programmes which are appropriate to children are watched during a learning phase. As the children get older, the boundaries are moved further and further which can be expressed, for example, in that the switch-off times of the controlled television are later in the day. After the learning phase, the checks as to how long the children may use the television and which programmes they may watch can be dispensed with. Here, for example, the external input variable can be the type of programme (news, sport, film) and/or a programme classification (for example for ages 16 years and above). Nowadays, such classifications or identifications are already being transmitted along with the programme signals. The time can again be used as an internal input variable. The current duration of use in the last 24 hours, for example, can be used as a feedback variable. The target variable is the switching on or switching off of the television. A time and classification-dependent use profile is produced.
When using a switching element for video recording control, the switching element learns which programmes are to be recorded and which not during a learning phase. Then, as a result of the information it has learned, it develops a use profile, whereupon all programmes are examined to see whether they correspond to the hitherto-selected programmes in its profile. In this way, laborious searching of confusing programme magazines can be dispensed with. Here, external input variables are the type of programme or the television channel, or also a classification of the programme. The target variable is a decision as to whether a programme should be recorded or not.
The switching elements, according to the invention, can also be used with Internet search engines, for example in order to select Internet pages provided or providable with identifications or classifications (for example, Internet pages can be provided with a classification by the user so that their contents are searched for keywords which are assigned to certain categories). The keyword search which is conventionally used with search engines has a range of disadvantages. Keywords or settings or specifications must be re-entered each time the search engine is used, automatic presentation of pages, especially taking a complex user desire profile into consideration, does not take place. In addition, the search for contradicting search terms proves problematic since each search term, for example, receives the same weighting during the search.
It is now possible, using the switching element according to the invention, to present an Internet user with those Internet pages which correspond to their interest or their usual search patterns. For example, if an Internet user calls up the Internet pages of a company, those documents concretely selected or read from the totality of the pages are classified (examples of classifications: culture, sport, business, ...).
Here, the concrete selection of a document corresponds to an external switching instruction. For example, the classification can take place by the entire document being searched for terms or keywords which are assigned to certain classifications or information headings.
Here, the quantity of classifications corresponds to the quantity of influence variables.
The non-selected documents (or a quantity or sub-quantity of these documents) are compared to the selected documents. A target function (0/1 function) is formed, wherein a selected document is assigned a 1, and a non-selected document is assigned a 0.
It can be specified, by means of such a classification of the selected documents, as to how far the classifications or influence variables, or their combinations, is relevant to a user. From this, it can be determined in which manner the classifications are to be weighted in a suitable manner. After an initialisation stage, the system is in a position to present the user with Internet pages or documents which are relevant on consideration of their previous search behaviour.
Switching elements according to the invention can be designed in a variety of variants. They can be realised as stand alone units or as series-connected hardware, i.e. connected in series, e.g. as a module, to a device.
Switching elements can also be designed as integrated hardware, for example as a hard-wired chip on a printed circuit board. It is also conceivable for switching elements to be designed as software. They can be active as threads e.g. on the operating system level as stand-alone processes or in sub-operating systems (JAVA system environment). As a software module attached in series, the switching element can be explicitly addressed by the user, i.e. the in- and outputs are assignable to individual operating systems or application programme functions by the user. As an integrated software module the switching element is not directly addressable, but its inputs and outputs are assigned in a fixed manner to operating systems or programme functions.
Finally, hybrid switching elements - i.e. software modules realised as hardware - are also conceivable, which are, for example, accessible to the user as chips or are assigned to the operating system or programme functions in a fixed manner.
It is conceivable to control the switching elements, . CA 02298460 1999-12-23 according to the invention, in such a way that user inputs are only possible during a learning phase so that in the following automatic phase the device equipped with the switching element is automatic. However, it is also possible to allow further inputs during such an automatic phase, for example to facilitate continual adaptation of the weighting of the influence variables.
Switching between the said phases could also take place automatically, for example on reaching a certain statistic level of previous inputs. Finally, it is conceivable to design an operating phase in which the functions of the learning phase and the automatic phase are realised simultaneously.
An automatic determination or adaptation of the chronological degree of fineness (i.e. the frequency of monitoring the estimation function) can take place by the minimum switching time unit being determined by detection of the relevant Fourier basic function.
It is also conceivable to provide an "evolutionary"
selection of relevant influence variables. For example, once it has become apparent over time that certain (pre-set or user-selected) influence variables actually have no influence on the switching behaviour, these can be recognised and extracted, and possibly (preferably by means of an automatic process) replaced by new influence variables. It is also conceivable, in a similar manner, to increase the number of influence variables over time.
It should be noted that the switch 10 can be designed as both an on/off element and also as a continual regulator (for example as a variable transformer).
It is also conceivable to feed the switching elements, according to the invention, back, i.e. to supply an output or switching variable of a switching element back to it as an input variable (for example via a relay).
In this way, regulator activities can be realised. In contrast to neuronal networks, the advantage results from the use of the switching elements, according to the invention, that chronologically-dependent processes can be detected and evaluated.