WO1999022464A1

WO1999022464A1 - Method and apparatus for selecting a batch of pending messages for a next transmission

Info

Publication number: WO1999022464A1
Application number: PCT/US1997/019711
Authority: WO
Inventors: Steven J. Goldberg
Original assignee: Motorola Inc.
Priority date: 1997-10-28
Filing date: 1997-10-28
Publication date: 1999-05-06

Abstract

Each batch of a plurality of batches of pending messages has a corresponding selective call protocol for use in transmitting the batch. A first memory (13) is arranged to store, for each of the plurality of batches, historical data, including an airtime efficiency rating of the selective call protocol corresponding to the batch, a latency measurement of the batch, and a message queue profile of the batch. A second memory (15) is arranged to store, for each of the plurality of batches of pending messages, current information including at least an associated priority of the batch, synchronization requirement of the corresponding protocol, and allowable transmission time for sending the batch. A sorter (11) is coupled to the first and second memories and is arranged to classifiy the historical data and the current information so as to select an optimum batch among the plurality of batches for the next transmission.

Description

METHOD AND APPARATUS FOR SELECTING A BATCH OF PENDING MESSAGES FOR A NEXT TRANSMISSION

Field of the Invention

This invention relates in general to messaging systems and more specifically to the use of contiguous logic in selecting a particular batch of pending messages for a next transmission in a messaging system.

Background of the Invention

Optimizing the selection of a batch of pending messages among a plurality of batches of pending messages for transmission of information presents a complex task that would probably become a computation nightmare to accurately model. Selective call systems such as those found in one and two way radio frequency (RF) paging networks, are typically of the store and forward type. This is an advantage allowing for the delay and reordering of information to efficiently pack the information before transmission. The processes involved are commonly referred to as batching and queuing. If there were no requirements other than efficiency of air time utilization and a single protocol, the process would be rather simple. In real systems however, other factors influence and complicate the process. Real systems will eventually require the ability to use multiple protocols and account for the numerous factors that make transmission of one batch of pending messages more advantageous than transmission of another batch. Thus, a need exists for an adaptive system and method for choosing a batch of pending messages among a plurality of batches of pending messages.

Summary of the Invention

A first aspect of the present invention is an apparatus for selecting, for a next transmission, a batch of pending messages among a plurality of batches of pending messages. Each of the plurality of batches has a corresponding selective call protocol for use in transmitting the batch. The apparatus comprises a first memory arranged to store, for each of the plurality of batches, historical data, comprising an airtime efficiency rating of the selective call protocol corresponding to the batch, a latency measurement of the batch, and a message queue profile of the batch. The apparatus further comprises a second memory arranged to store for each of the plurality of batches of pending messages, current information comprising at least an associated priority of the batch, synchronization requirement of the corresponding protocol, and allowable transmission time for sending the batch. The apparatus also includes a sorter coupled to the first and second memories and arranged to classify the historical data and the current information so as to select an optimum batch among the plurality of batches for the next transmission.

A second aspect of the present invention is a method for selecting, for a next transmission, a batch of pending messages among a plurality of batches of pending messages. Each of the plurality of batches has a corresponding selective call protocol for use in transmitting the batch. The method comprises the steps of identifying a plurality of inputs for use in selecting the batch, and defining contiguous logic sets for each of the plurality of inputs identified. The method further comprises the steps of defining a contiguous logic set for a desired output, and defining rules relating the contiguous logic sets for each of the plurality of inputs with the contiguous logic set for the desired output. The method also includes the steps of maintaining statistics based on information received from the plurality of inputs, and executing the rules when a selection of one of the plurality of batches is required based on the statistics maintained. In addition, the method includes the step of selecting the batch with a highest need as determined by the rules.

Brief Description of the Drawings

FIG. 1 is an adaptive system for choosing a batch of pending messages in accordance with the present invention.

FIG. 2 is a plot of a membership function for an Efficiency characteristic of a selected protocol in accordance with the present invention.

FIG. 3 is a plot of a membership function for a Latency characteristic of a selected batch of pending messages in accordance with the present invention. FIG. 4 is a plot of a membership function for a Pending Buffer Utilization characteristic of a selected batch of pending messages in accordance with the present invention.

FIG. 5 is a plot of a membership function for a Page Pending Trend characteristic of a selected batch of pending messages in accordance with the present invention.

FIG. 6 is a plot of a membership function for a Priority characteristic of a selected batch of pending messages in accordance with the present invention. FIG. 7 is a plot of a membership function for FLEX Sync Required in accordance with the present invention.

FIG. 8 is a plot of an output function for a Transmit Need characteristic of a selected batch of pending messages in accordance with the present invention. FIG. 9 is a linguistic rule-base in accordance with the present invention.

FIG. 10 is a flow chart of a method of choosing a batch of pending messages in accordance with the present invention.

Description of the Preferred Embodiment

Most contiguous logic systems use a rule-base as their central structure. Rules, typically cast in an if-then syntax, represent system operation, mapping inputs to outputs. Measured crisp input values are "fuzzified", using membership functions, into fuzzy truth values or degrees of membership. These are then applied as conditions to the rules contained in the rule-base, with triggered rules specifying necessary actions, again as fuzzy truth values. These actions are combined and "defuzzified" into crisp, executable system outputs. Where inputs and outputs are continuous (as in control applications), this fuzzify-infer-defuzzify process is performed on an ongoing basis, at regular sampling intervals. In summary, crisp inputs are measured and assigned fuzzy membership values as part of the fuzzification step, which are applied as conditions to the rules in the rule-base. The rules then specify what actions are to be taken, although in fuzzy terms. Fuzzy actions from typically two or more rules are combined and transformed back into executable system outputs. Conceptually this process is similar to the use of a Fast Fourier Transform and its inverse to transform time domain signals into the frequency domain, to process the resulting frequencies, and then to transform the results back into the time domain. The added expense of transforming between time and frequency domains is justified because the system model is easier to understand and manipulate in terms of frequencies.

Similarly, a contiguous logic system "transforms" signals from the "crisp domain" to the "fuzzy domain", makes decisions based on these fuzzy values and knowledge of the desired system operation cast in fuzzy terms (rules), and then transforms the results back into the crisp domain for execution. The justification is, as with frequency domain processing, that the system model is easier to understand and manipulate in the fuzzy domain than in the crisp domain. This basic fuzzy rule-based structure can be used in choosing a batch of pending messages among a plurality of pending messages. By allowing for flexibility in the definition of fuzzy logic operators, and especially in how action combination/defuzzification is performed, the breadth of application is even further increased. Referring to FIG. 1, an adaptive system 10 for choosing a batch of pending messages is shown in accordance with the present invention. The system 10 includes a system controller 12 or radio frequency manager that uses a contiguous logic controller 11 or sorter to determine which batch of pending messages to use. The contiguous logic controller 11 or sorter preferably comprises a fuzzifier 14, a rule-based element 16 or inference engine and a defuzzifier 18. The sorter 11 preferably classifies historical data and current information so as to select an optimum batch of pending messages among a plurality of batches of pending messages available for the next transmission of the selective call messages and, alternatively, to select a second optimum batch for the next transmission if the optimum batch is not able to be sent for any reason. Preferably, the fuzzifier 14 generates membership values for historical and current input conditions while the rule-based element 16 applies a set of predetermined rules of the current input conditions to perform a mapping with the historical conditions. The defuzzifier 18 preferably generates a non-fuzzy predicted recommendation from the mapping, and selects the optimum batch or, alternatively, selects a second optimum batch. The contiguous logic controller or sorter 11 simplifies the decision making process for the system controller 12 in determining whether to choose one batch or another for transmission of a message through a transmitter 20. The sorter 11 uses the fuzzifier 14 as an input processing element for receiving biases for airtime efficiency, latency, priority, and battery savings among other biases. The sorter 11 uses the defuzzifier 18 as an output processing element for outputting the optimum batch and finally, the rule-based element 16 is used as an intermediate processing element for coupling the input processing element with the output processing element. Additionally, the system 10 includes an operator to modify the rule-based element 16 and further preferably provides for built-in limits to prevent modifications from exceeding safeguard limits. The modifications can be achieved using a graphical input such as the interface 17 shown in FIG. 1. In the examples presented, the assumption is made that the system controller 12 needs to choose between only two protocols for simplicity. The paging protocols used in the examples shown are FLEX™ and POCSAG, but other protocols in any combination could potentially be used such as pACT, APOC, ERMES, or GOLAY. In addition, for further simplicity, the assumption is made that only two batches of pending messages are maintained—one batch for FLEX and the other batch for POCSAG. Thus "selecting a batch" is equivalent to "selecting a protocol" for the simple examples used herein. It will be appreciated that a large number of alternative examples can be envisioned, as well. For example, there could be a High Priority batch and a Low Priority batch for each protocol; there could be multiple batches for each protocol for pagers utilizing different battery saving intervals, and so on. For these latter examples, "selecting a protocol" is not equivalent to "selecting a batch," because a single protocol is used for the transmission of multiple different batches.

The radio frequency manager or system controller 12 preferably comprises a first memory 13 for storing, for each of the plurality of batches, historical data comprising an airtime efficiency rating of the selective call protocol corresponding to the batch, a latency measurement of the batch, and a message queue profile of the batch. Of course, other historical data can be tracked as suitably required. The radio frequency manager or system controller 12 also preferably comprises a second memory 15 arranged to store for each of the plurality of batches of pending messages, current information comprising at least an associated priority of the batch, synchronization requirement of the corresponding protocol, and allowable transmission time for sending the batch.

FIGs. 2-7 show plots of membership functions for different characteristics corresponding to a selected batch of pending messages or protocol therefor, the characteristics used in the fuzzifier 14 to determine a degree of membership for the selected batch of pending messages in defined sets of each of the characteristics, given a crisp input value. Measured crisp input values are "fuzzified", using membership functions, into fuzzy truth values or degrees of membership. For instance, a FLEX message having a measured efficiency of 65%, would have a degree of membership of 0.15 in the set "HIGH" while also having a degree of membership of 0.27 in the set "LOW" in accordance with the membership function shown in FIG. 2. These are then applied as conditions to the rules contained in the rule-base of FIG. 9, with triggered rules specifying necessary actions, again as fuzzy truth values. For instance, see rule stating "IF flex_latency IS high AND flex_efficiency IS high THEN flex_xmit_need IS high." This rule and other rules are combined and "defuzzified" into crisp, executable system outputs.

More specifically, there are a number of paging protocols each having different characteristics which can influence their efficiency, and the efficiency of the selective call receivers they are signaling. These protocols often must coexist on the same network, with time allocated for each one's operation in a manner which optimizes its particular requirements. Message batches and their associated transmission protocols have in common among other things the factors of airtime efficiency, latency, priority, and subscriber unit battery savings. Each of these factors is further described below.

Airtime Efficiency is a key factor for service providers since servicing as large a volume of paging traffic as possible will maximize their profits. If service providers charge flat rates per subscriber, this means the service providers must support as many pages per unit of RF bandwidth as possible. If they charge by the packet of data sent, the raw number of packets becomes significant. FIG. 2 is a plot of a membership function reflecting an Efficiency characteristic of a selected protocol in accordance with the present invention.

With respect to latency, pages or messages can become old and stale. Those sent with the idea of immediately conveying information, can be of decreasing value as they are delayed in the system. In some cases this can cause additional traffic in a system as people reinitiate a page, believing a page has been missed or ignored, when in fact it may not have left the system yet. In extreme cases (examples: medical & fire) high latency can actually be life threatening. FIG. 3 is a plot of a membership function reflecting a Latency characteristic of a selected batch of pending messages in accordance with the present invention.

With respect to priority, the relative importance of a page can be entered into a system as a priority level. FIG. 6 is a plot of a membership function reflecting a Priority characteristic of a selected batch of pending messages in accordance with the present invention. Messages with higher priorities should in general be sent before messages with lower priorities. This may not always happen, however, even in efficient systems. If a particular priority page is batched with other pages to be sent, a lower priority page may be blocked temporarily, while still lower priority pages which are not blocked are sent. A preemptive priority page is a special case causing most other paging rules to become invalid until the page is sent.

Battery savings at a subscriber unit 22 (shown in FIG. 1) is a key advantage of pagers over other communication devices due to the resultant long useful battery life. This is mostly achieved by making use of the store and forward nature of paging systems, to allow pagers to 'sleep' until specific prearranged times. Different paging protocols handle this feature in different ways. The FLEX™ paging protocol, for example, sends pagers synchronization transmissions at a minimum rate even when there are no messages to be transmitted. If the pagers do not detect this sync signal when they awaken, they must search for it. Searching in turn requires the pager to remain awake longer then normal, which consumes more battery power. Thus, synchronization is critical in the FLEX™ paging protocol for retaining certain battery saving advantages. FIG. 7 is a plot of a membership function for FLEX Sync Required in accordance with the present invention.

Another factor is the relative state of the system. During low RF utilization periods, efficiency may not be an important factor. When the system becomes heavily loaded however, efficiency may become a more important issue. These variables are often optimized by different approaches to batching and queuing. A strategy which is good for one variable can be detrimental to one or more of the others. As a paging system matures, the mix of pager types with respect to protocols in use, as well as pager types (numeric, alphanumeric, voice, tone only, etc.) may vary.

Considering all the above tradeoffs in an all inclusive, adaptive manner is a nontrivial task. The exact definitions and relative importance of factors can vary amongst customers. In fact, a service provider can find it difficult to define his own biases given the need to satisfy all customer requirements, regardless of conflicting relationships among the requirements.

Over time the situations considered optimum also change. This can be because of better understanding of the system's operation and /or changing customer needs. Prior art page batching and queuing management systems have been optimized towards one type of scenario, which has made customization difficult.

The ideal system would take into account all the previous requirements in a form that is easy to understand and adjust, yet is robust and extensible. The present invention can closely approach the ideal system when sufficient inputs are used. The exemplary embodiment disclosed considers operation with only two protocols: POCSAG and FLEX™. Other embodiments are easily extensible to an arbitrary number of different protocols.

In a real example, statistics relating to the batching and queuing of each protocol can generate inputs for the decision making process. A potential set preferably includes efficiency, latency, pendency (FIG. 4), pendency trend (FIG. 5), priority, and synchronization. Given the protocol's recent historic use of airtime, statistics on efficiency will show the fraction of time used for actual data transmission. Statistics on latency will show the average length of time pages are remaining in the queues before being sent. Statistics on pendency will show the number of pages actually remaining in the queues. Statistics on a pendency trend is derived from the number of messages pending during a specified time interval. A figure of merit indicating the increase and reduction in the size of the queue is therefore derived. Statistics on priority will show the priority assigned to a page either at input, or incremented by the system because of excessive individual residence in a queue. Finally, a mixed system using the FLEX™ protocol must account for the system's requirement for synchronization for the FLEX™ protocol.

In the prior art, accounting for these variables has been very straightforward. If FLEX™ synchronization was required, the FLEX™ protocol was used. Otherwise, if either protocol had a higher priority item pending, the higher priority item was sent using the corresponding protocol. Otherwise, the protocol with the oldest or most pages pending was chosen.

Using the approach of the present invention requires defining contiguous logic sets for each of the inputs shown in FIGS. 2-7. Next, a contiguous logic set for a "need" to transmit a particular batch of pending messages is defined which serves as an output set as shown in FIG. 8. Rules are then defined relating the inputs to the output set or sets. Preferably, the rules are easily defined linguistically as shown in FIG. 9. The rules are then executed based on the statistics as they exist at each batch choosing. Whichever batch has the higher "need" determines the choice of the batch to be transmitted next. Alternatively, the rules can be defined in other forms other than linguistically. For instance, the rules could be defined as a matrix of sets or the rules could utilize a neural net (see Neural Networks and Fuzzy Systems, by Bart Kosko 1992) to refine the rules as the system operates in a real setting. Contiguous logic systems use neural systems to learn fuzzy rules from examples or to tune the rules. The net learns the fuzzy rules by adapting its dynamic structure. The rules emerge as the equilibrium states of the neural dynamic system. One of ordinary skill in the art will appreciate that the embodiments described and the linguistic rules disclosed are merely exemplary and that variations and modifications can be made within the scope of the present invention as defined by the appended claims.

FIG. 10 illustrates a flow chart of a method 100 of choosing a batch of pending messages among a plurality of batches of pending messages in accordance with the present invention. At step 102, a plurality of inputs are identified for use in a choosing a batch of pending messages including inputs for airtime efficiency, latency, message queue profile, and subscriber unit battery savings profile. At step 104, contiguous logic sets are defined or created for each of the inputs identified. At step 106, a contiguous logic set for the desired output is created. Then, at step 108, rules are defined relating the contiguous logic sets for each input with the contiguous logic set for the desired output. At step 110, statistics are maintained based on information received from the inputs. The rules are executed at step 112 when a choice of one of the plurality of batches of pending messages is required based on the statistics maintained. Finally, at step 114 the batch of pending messages is selected with the highest need as determined by the rules. As a consequence, at step 116, information is transmitted using the batch of pending messages selected to have the highest need. Optionally, the method 100 could further comprise the step 118 of modifying the relationships between the contiguous logic sets for each input and the contiguous logic set for the desired output by defining an inference engine using a linguistic rule base.

The advantages of this approach are many. Using contiguous logic allows all concerns regarding selection of the batch of pending messages to be addressed. New rules can be added rapidly, since the decision factors are mostly of a parallel calculating nature. Transitions from one scenario to another is gradual rather than abrupt. Further, this prevents one protocol from completely dominating another excessively, when a partial bias will balance the system in a reasonable manner.

If desired, additional protocols can be added to the consideration by defining similar rules, which linguistically change the existing ones.

Finally, the system manager can be given control over the weights of the rules. This allows flexibility in biasing the system's operation, while minimizing the possibility of totally disrupting the system.

In summary, the present invention preferably uses contiguous logic to define the decision making variables, wherein there can be consideration of all the variables affecting the decision process at all times. A customer such as a service provider can access, define, or refine weights which allow biasing of the decision making process, without leading to undesirable operating characteristics. Finally use of linguistic rules can be used to determine a near optimum decision.

Claims

1. An apparatus for selecting, for a next transmission, a batch of pending messages among a plurality of batches of pending messages, wherein each of the plurality of batches has a corresponding selective call protocol for use in transmitting the batch, the apparatus comprising: a first memory arranged to store, for each of the plurality of batches, historical data, comprising an airtime efficiency rating of the selective call protocol corresponding to the batch, a latency measurement of the batch, and a message queue profile of the batch; a second memory arranged to store, for each of the plurality of batches of pending messages, current information comprising at least an associated priority of the batch, synchronization requirement of the corresponding protocol, and allowable transmission time for sending the batch; and a sorter coupled to the first and second memories and arranged to classify the historical data and the current information so as to select an optimum batch among the plurality of batches for the next transmission.

2. The apparatus as set forth in claim 1, wherein the sorter is arranged to alternatively select a second optimum batch for the next transmission if the optimum batch is not available.

3. The apparatus as set forth in claim 1, wherein the apparatus further comprises a plurality of processing elements including an input processing element for receiving bias settings for airtime efficiency, latency, priority, and battery savings; an output processing element for outputting the optimum protocol identified; and an intermediate processing element for coupling the input processing element with the output processing element.

4. The apparatus as set forth in claim 1 wherein the sorter comprises: a fuzzifier for generating membership values for historical conditions and current input conditions; and a rule base for applying a set of predetermined rules to the current input conditions to perform a mapping with the historical conditions.

5. The apparatus as set forth in claim 4 wherein the sorter further comprises: a defuzzifier for generating a non-fuzzy prediction recommendation 5 from the mapping, and for selecting the optimum batch and alternatively, for selecting the second optimum batch.

6. The apparatus as set forth in claim 4, wherein the sorter further comprises an operator to modify the rule base for a fuzzy logic rules o engine and a limiter to prevent such modifications from causing safeguard limits to be exceeded.

7. The apparatus as set forth in claim 6, wherein a linguistic input is used to modify relationships defining an inference engine operation associated 5 with the rule base for the fuzzy logic rules engine.

8. A radio frequency manager for selecting, for a next transmission, a batch of pending messages among a plurality of batches of pending messages, wherein each of the plurality of batches has a corresponding 0 selective call protocol for use in transmitting the batch, the radio frequency manager comprising: a first memory arranged to store, for each of the plurality of batches, historical data comprising an airtime efficiency rating of the selective call protocol corresponding to the batch, a latency measurement of the 5 batch, and a message queue profile of the batch; a second memory arranged to store, for each of the plurality of batches of pending messages, current information comprising at least an associated priority of the batch, synchronization requirement of the corresponding protocol, and allowable transmission time for the 0 sending the batch; and a sorter coupled to the first and second memories and arranged to classify the historical data and the current information using fuzzy logic so as to select an optimum batch for the next transmission, and, alternatively, to select a second optimum batch for the next 5 transmission of selective call messages if the optimum batch is not available, wherein the sorter comprises a fuzzifier for generating membership values for historical conditions and current input conditions, a rule base for applying a set of predetermined rules to the current input conditions to perform a mapping with the historical conditions, and a defuzzifier for generating a non-fuzzy prediction recommendation from the mapping, and for generating the optimum batch and, alternatively, for generating the second optimum batch; a plurality of processing elements including an input processing element for receiving bias settings for airtime efficiency, latency, priority, and battery savings; an output processing element for outputting the optimum batch identified; and an intermediate processing element for coupling the input processing element with the output processing element.

9. A method for selecting, for a next transmission, a batch of pending messages among a plurality of batches of pending messages, wherein each of the plurality of batches has a corresponding selective call protocol for use in transmitting the batch, comprising the steps of:

a) identifying a plurality of inputs for use in selecting the batch; b) defining contiguous logic sets for each of the plurality of inputs identified; c) defining a contiguous logic set for a desired output; d) defining rules relating the contiguous logic sets for each of the plurality of inputs with the contiguous logic set for the desired output; e) maintaining statistics based on information received from the plurality of inputs; f) executing the rules when a selection of one of the plurality of batches is required based on the statistics maintained; and g) selecting the batch with a highest need as determined by the rules.

10. The method of claim 9, further comprising the step of h) transmitting the batch selected to have the highest need.