CN112290981A - Parallel feedback communication method based on multiple-input multiple-output wireless charging system - Google Patents

Abstract

A parallel feedback communication method based on a multiple-input multiple-output wireless charging system comprises a collision perception parallel decoding scheme: RXs, which will respond to the TX analog ping concurrently and feed back their necessary data, in the proposed MRC-WPT system, RX encodes its data by switching the oscillator circuit to two possible states.

Description

Parallel feedback communication method based on multiple-input multiple-output wireless charging system

Technical Field

The invention relates to a parallel feedback communication method based on a multi-input multi-output wireless charging system, and belongs to the field of digital wireless charging and wireless communication.

Background

Wireless charging technology has been widely developed and used in recent years, and various wireless charging products have come into the field of view of consumers. The reasons for this are firstly caused by the demand and secondly technical applicability. The requirement is that a large number of wireless power devices enter thousands of households since the 21 st century, such as mobile phones, tablets, bracelets, wireless keyboards and mice, in order to supply power to the devices, traditional wired charging is cumbersome, and users often need to consider the electric quantity of the devices; the wireless charging realizes wireless energy transmission, so that the charging and discharging become possible, meanwhile, the wireless electric energy transmission is more beneficial to the closed design of products, and the use safety is improved. These reasons have together contributed to the development of wireless charging technology.

Wireless charging products in the current market mainly fall into two categories, namely a wireless high-power transmission system for supplying power to an automobile and a wireless charger for small-electric-quantity equipment (a mobile phone, a bracelet and the like). However, from the point of view of the usage in the market, the current wireless charging has a great limitation, and the most important problem is represented by the wireless charging distance. For a wireless charger of a low-power device, almost all known products need to put an electric appliance on the wireless charger according to the position, and the farthest charging distance is generally lower than 1 cm. The wireless charger is not consistent with the definition of wireless charging in terms of quantity, so that the user is only free from the trouble of plugging and unplugging the connector, and the use and popularization of wireless charging are limited.

Aiming at the hard defect, researchers have intensively studied Wireless Power Transmission (Wireless Power Transmission) systems in recent years, and a magnetic resonance technology [1] and a Multiple Input Multiple Output (MIMO) technology [2] are applied to Wireless charging to solve the problems of too short distance, too small Power and the like.

However, the wireless charging system under MIMO system has a major problem, and there is a core problem in channel estimation of receiving end communication and magnetic field energy transmission. In a magnetic resonance-based multi-input multi-output wireless charging system, the feedback of a receiving end and the estimation of an energy transmission channel are very important for the environment sensing capability of the system, equipment can be powered as required by realizing transmitting and receiving communication and measuring and calculating the channel [3], and meanwhile, a reliable channel coefficient is favorable for better beam forming [4 ]. At present, the solution of the MRC-WPT system based on MIMO is neither efficient nor mature, and most of the existing researches are focused on the optimization algorithm of the current (current) adjustment of a transmitting terminal and adopt a simple feedback communication and channel estimation scheme. Channel estimation schemes that involve omitting communication and ignoring differences in energy requirements at the receiving end either simply assume the existence of a communication link or a priori known channel conditions or use an underlying approach to achieve one-to-one communication at both the transmitting and receiving ends.

In our work, it was found that there is a myriad of connections between the current set at the transmitting end and the switch state clusters at the receiving end. Based on the phenomenon, a parallel multi-order decoding scheme which relies on the recognition of the joint switch state of each cluster is provided, and in the state clustering stage, a two-layer clustering mechanism is introduced to solve the 'dominant RXs' challenge which causes inaccurate classification results; in the cluster identification phase, we solve the challenge of "fuzzy identification candidates" by calibration based on energy transfer channel estimation. Finally, we have made a test prototype using the existing materials and achieved the desired effect.

Disclosure of Invention

The invention solves the problem of feedback communication of the receiving end in a wireless charging system, identifies the state of multiple receiving ends and calculates the corresponding energy transmission channel, and the invention aims to realize the following technical scheme: a parallel feedback communication method based on a multiple-input multiple-output wireless charging system comprises the following steps:

step 1: collision-aware parallel decoding scheme

RXs will respond concurrently to TX emulated pings and feed back their necessary data, in the proposed MRC-WPT system, RX will switch the tank circuit to two possibleThe states encode their data, "open (O)" and "closed (S)". When Q RXs are transmitted simultaneously, the collision signal will have 2^QA combined state, coordinate coordination is carried out among all TXs by measuring TX current, a combined state of RXs is identified, mutual inductance coefficients of TXRX/RX-RX are estimated, TX voltage is fixed to a constant value, and TX current is used as a measuring variable,

step 2: observation of phenomena

Clustering phenomenon-the presence of RXs affects the TX current value given that the TX voltage is constant. We performed experiments on our MRC-WPT prototype test bench using 2 TXs and 2 RXs, we fixed the TX input voltage to 8V and switched the RX oscillating circuit to an open state, implemented RX feedback communication, collected and plotted the TX current value, which was significantly affected by the RX on-off state switching.

And step 3: construction of single hop transition graphs

Based on the symbols collected in the n-dimensional space (receiving end switch state representation), we construct a single-hop transition Graph (OFG), in which the vertex set contains 2^QA cluster of symbols and an edge set contains any pair of adjacent edges.

According to a circuit equation, the cluster distribution of the MIMO MRC-WPT system is more regular than that of the RFID communication. When dominant RXs of K ≦ Q is present and its effect on TXs is much greater than the other RXs, 2^QThe clusters tend to form 2^KEach group consisting of 2^Q-KAnd (4) cluster composition. The classifier may not be able to distinguish between clusters that are distributed tightly within a group, resulting in incorrect classification results. As shown in fig. 4(a), each group has 4 groups, each group has 2 clusters, and the classifier has a result of a judgment error.

To improve the accuracy of classification, we use a two-level clustering mechanism, at the first level, we try to locate 2^KWhen there are a sufficient number of symbols, the distribution of symbols in all clusters is normal according to the circuit equation. Thus, the parameter K of the first layer can be adjusted to achieve an even distribution of symbols among groups.

On the second layer, we further fractionate each fractionTo obtain 2^Q-KAnd (4) clustering. FIG. 4(c) shows the second layer clustering results of the upper left group of FIG. 4 (b). The LDBC algorithm introduced therein is used to locate the center of the cluster in both the first and second layers.

After determining the center of each cluster, we will perform a reliable symbol allocation. A symbol is classified as a cluster if the probability (determined by the distance of the symbol position from the cluster center) of the symbol in the spatial domain is greater than a given threshold. For confusing symbols that may exist in multiple clusters of overlapping spaces, the classification is based on a joint consideration of the probabilities in the space and time domains, where the probability in the time domain is determined by neighboring symbols within the same time window. After symbolic clustering, the next step is to build connections in the graph. Based on the observation that the transition probability between neighboring clusters is significantly higher than the transition probability between non-neighboring clusters, we can identify neighboring clusters from the transition probabilities.

And 4, step 4: potential cluster identification

RFID communication-related research proposes a layer-based cluster identification algorithm that starts with a particular anchor cluster, assuming its combined state is known. This assumption is reasonable in the RFID scenario because the charged tags do not affect the signal received in the IQ domain, and when all tags are charged, clusters representing all "low" states can be identified.

And 5: calibration based on channel estimation

For any given cluster identification candidate, we can obtain the mutual inductance of TX-RX and RX-RX, respectively, from clusters with one or two "closures" RXs. After all the mutual inductances are known, we further evaluate the superiority and inferiority of a given cluster identification by comparing the expected and measured TX currents. The reason for this is that the mutual inductance obtained by the different clusters will be consistent with other clusters in the effective cluster recognition and will conflict with other clusters in the ineffective cluster recognition. Therefore, we will select the best cluster identification by calibration based on channel estimation.

Step 6: decoding

Finally, we can perform decoding based on the best cluster identification, where each RXs combined state of opening and closing is identified. By examining these combined states, the scheme outputs a sequence of "open" and "closed" to represent the signal transmitted by each RX. In collision scenarios, an error label during symbol clustering may increase the error rate of a single RX. The standardized FM0 and Miller predictable state flipping patterns present a possible solution to correct errors in the state sequence before it is input to the decoder.

The invention aims to establish a receiving feedback communication mechanism and estimate and measure an energy transmission channel for a multi-input multi-output wireless charging system (MIMO MRC-WPT) based on magnetic resonance, thereby achieving the environment perception capability of the MIMO MRC-WPT system.

In the present invention, we have studied high performance feedback communication and channel estimation schemes, and in order to save power consumption and hardware cost, we consider the in-band communication link, and avoid unnecessary RX part changes by assuming that the workflow of RX operation and the design of the oscillation feedback circuit both meet the Qi standard specification.

Drawings

The receiver design in the Qi protocol of fig. 1.

Fig. 2 Transmit (TX) current clustering.

Figure 3 system workflow.

FIG. 4 is a two-tier clustering mechanism.

Fig. 5 system prototype architecture.

Fig. 6 shows a test prototype pictorial representation.

Detailed Description

In the following, with reference to the detailed contents of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described, and our work focuses on a general MIMO MRC-WPT system including N TX and Q RX, since the end inductive reactance and the capacitive reactance including the transmitting end and the receiving end will cancel each other at the resonant frequency due to the application of the magnetic resonance technology, we can ignore the terms related to the self-inductance and the capacitance in the circuit equation.

Circuit formula

We will want to

And

shown as complex steady state currents in the nth TX and qth RX coils, respectively. Applying kirchhoff's circuit law to obtain a circuit equation of RX q as follows:

complex source voltage of TX n, noted

From the TX/RX current, it can be found by:

for more convenient representation, we can express the above formula in a matrix form.

Receiving end oscillating circuit

In the Qi specification, the operation of a wireless charging system is divided into two main phases, an initialization phase and a power transfer phase. During the initialization phase, the RX will respond to the analog ping from the TX with one signal strength packet. If the signal strength packet is valid, the TX will recognize it and collect the configuration and setup information provided by the RX. Thus, the system enters a power transfer phase to charge the load. When the power transfer phase is over, the system returns to the initialization phase. The receiving end circuit and three different switching modes are shown in fig. 1. A parallel feedback communication method based on a multiple-input multiple-output wireless charging system comprises the following steps:

step 1: collision-aware parallel decoding scheme

The invention mainly researches a concurrent RX feedback communication scheme and simultaneously carries out channel estimation. More specifically, we assume that multiple Qi specifications compatible with RXs will respond concurrently to a TX analog ping and feed back their necessary data, such as RX load resistance and power requirements.

In the proposed MRC-WPT system, RX encodes its data by switching the tank circuit to two possible states, "open (O)" and "closed (S)". When Q RXs are transmitted simultaneously, the collision signal will have 2^QAnd a merge state. Although circuit equation (3) gives the TX current/voltage versus RX state, we still cannot directly derive the RX state due to the unknown TX-RX/RX-RX mutual inductance.

The scheme is performed at the TX end, coordinate coordination is performed among all TXs by measuring TX current, the combined state of RXs is identified, and the mutual inductance coefficient of TXRX/RX-RX is estimated. Here we fix the TX voltage to a constant value and use the TX current as a measured variable, since it is more convenient to control the voltage than the current in real circumstances. Also, we assume that the mutual inductance of TX-TX is known and can be measured off-line.

Step 2: observation of phenomena

Clustering phenomenon As can be seen from the circuit equation (3), the existence of RXs affects the TX current value given that the TX voltage is not changed. We performed experiments on our MRC-WPT prototype bench using 2 TXs and 2 RXs. We fix the TX input voltage to 8V and switch the RX tank circuit to an open state, enabling RX feedback communication. TX current values were collected and plotted as shown on the left of fig. 2. As shown, we can notice that the switching of RX on and off states has a significant effect on TX current value.

To show more clearly, we further plot the sample points in two-dimensional coordinates, where the two axes are the current amplitudes of the two TXs, respectively. As shown on the right of fig. 2, the sample points were clearly clustered into four kinds, which completely correspond to the four combined switch states of the two kinds RXs. From the observed clustering, there is an opportunity to identify the RX state by clustering the measured TX currents.

We note that in related academic research, similar clustering phenomena have been discussed and used for parallel decoding of RFID communications. Their solution is to perform symbol clustering on a two-dimensional IQ plane and then perform single-hop transition graph (OFG) based decoding to identify the combined state of each cluster. They assume that the state flips of different RFID tags are staggered most of the time due to inaccurate tag synchronization. Therefore, typically only one RX state flips at a time, and the transition probability between neighboring clusters is much higher than that between non-neighboring clusters. They also assume that the combined state of the anchor is known before receiving information from the RFID tag because all RXs are in the off-state before the feedback communication. Our basic idea is to extend the application of a similar method to the MIMOMRC-WPT system, whose workflow is shown in figure 3. First, we need to switch RXs switch states to get TXs measured series of current data.

And step 3: construction of single hop transition graphs

Based on the symbols collected in the n-dimensional space (receiving end switch state representation), we construct a single-hop transition Graph (OFG), in which the vertex set contains 2^QA cluster of symbols and an edge set contains any pair of adjacent edges.

According to a circuit equation, the cluster distribution of the MIMO MRC-WPT system is more regular than that of the RFID communication. When dominant RXs of K ≦ Q is present and its effect on TXs is much greater than the other RXs, 2^QThe clusters tend to form 2^KEach group consisting of 2^Q-KAnd (4) cluster composition. The classifier may not be able to distinguish between clusters that are distributed tightly within a group, resulting in incorrect classification results. As shown in fig. 4(a), each group has 4 groups, each group has 2 clusters, and the classifier has a result of a judgment error.

To improve the accuracy of classification, we adopt a double-layer clustering mechanism. In the first layer, we tasteTrial position 2^KGroup, as shown in FIG. 4 (b). When there are a sufficient number of symbols, the distribution of symbols in all clusters is normal according to the circuit equation. Thus, the parameter K of the first layer can be adjusted to achieve an even distribution of symbols among groups.

At the second level, we further classified each group to obtain 2^Q-KAnd (4) clustering. FIG. 4(c) shows the second layer clustering results of the upper left group of FIG. 4 (b). The LDBC algorithm introduced therein is used to locate the center of the cluster in both the first and second layers.

After determining the center of each cluster, we will perform a reliable symbol allocation. A symbol is classified as a cluster if the probability (determined by the distance of the symbol position from the cluster center) of the symbol in the spatial domain is greater than a given threshold. For confusing symbols that may exist in multiple clusters of overlapping spaces, the classification is based on a joint consideration of the probabilities in the space and time domains, where the probability in the time domain is determined by neighboring symbols within the same time window. After symbolic clustering, the next step is to build connections in the graph. Based on the observation that the transition probability between neighboring clusters is significantly higher than the transition probability between non-neighboring clusters, we can identify neighboring clusters from the transition probabilities.

And 4, step 4: potential cluster identification

RFID communication-related research proposes a layer-based cluster identification algorithm that starts with a particular anchor cluster, assuming its combined state is known. This assumption is reasonable in the RFID scenario because the charged tags do not affect the signal received in the IQ domain, and when all tags are charged, clusters representing all "low" states can be identified.

However, this assumption cannot be guaranteed in the proposed MIMO MRC-WPT system. RXs in the charged state still affects TXs current according to the circuit equation. Therefore, we cannot consider the measured data of the charging phase as a "fully open" combined state (i.e., a state where all RXs are open). We can select the cluster closest to the "fully on" state theoretical center as the anchor cluster, but due to the clustering grouping phenomenon mentioned earlier, and considering noise and measurement errors, selection based on the closest distance may produce errors. Therefore, we select the top K nearest clusters as potential "full on" clusters. The number K may be adjusted according to the density of the cluster distribution. For example, we can select K as the number of clusters within a group.

With these potential "full on" anchor clusters, we can use cluster recognition algorithms to obtain corresponding cluster recognition candidates. Furthermore, since the algorithm is based on likelihood, we may have multiple candidate identifications even given a "fully-open" anchor cluster. The final anchor cluster is determined by calibration based on the energy transfer channel estimate.

And 5: calibration based on channel estimation

For any given cluster identification candidate, we can obtain the mutual inductance of TX-RX and RX-RX, respectively, from clusters with one or two "closures" RXs. After all the mutual inductances are known, we further evaluate the superiority and inferiority of a given cluster identification by comparing the expected and measured TX currents. The reason for this is that the mutual inductance obtained by the different clusters will be consistent with other clusters in the effective cluster recognition and will conflict with other clusters in the ineffective cluster recognition. Therefore, we will select the best cluster identification by calibration based on channel estimation.

Step 6: decoding

Finally, we can perform decoding based on the best cluster identification, where each RXs combined state of opening and closing is identified. By examining these combined states, the scheme outputs a sequence of "open" and "closed" to represent the signal transmitted by each RX. In collision scenarios, an error label during symbol clustering may increase the error rate of a single RX. The standardized FM0 and Miller predictable state flipping patterns present a possible solution to correct errors in the state sequence before it is input to the decoder.

The invention carries out deep research on two important problems of the MIMO MRC-WPT system, namely parallel in-band feedback communication and measurement and calculation of an energy transmission channel.

The invention introduces two challenges with respect to two problems, applies an extended algorithm in the field of RFID communication, solves the 'dominant RXs' challenge using a two-layer classification mechanism, and solves the 'fuzzy recognition candidate object' challenge using a channel estimation calibration mechanism.

In an actual test platform, the invention obtains the state identification accuracy rate of over 97% under all test conditions, and the effective communication distance reaches 45cm, which shows the effectiveness of the system concurrent feedback communication; meanwhile, simulation experiments prove that the difference between the transmission efficiency of the system and the optimal beamforming parameter is not more than 5% under most conditions (except that the number of the transmitting ends is far less than that of the receiving ends), so that the reliability of the energy transmission channel parameter estimation scheme is ensured.

The prototype architecture of the system is shown in fig. 5.

Examples of the embodiments

The application scenario is shown in fig. 6, where fig. 6 includes all the components of the present invention, and 2 transmitting terminals and 3 receiving terminals are used. The TXController is a computer and is used for receiving the measurement data uploaded by the transmitting end, performing complex matrix operation and giving out the optimal beam forming parameters according to the estimated energy transmission channel; the RXController is used for controlling the on-off state of the receiving end.

In the system, the resonant frequencies of all the transmitting ends and the receiving ends are 1.0MHz, which is in the frequency range of the common wireless power transmission system and can not interfere with other wireless device frequency bands. In the test prototype object of fig. 6 to which the present invention was applied, the coil of the receiving terminal was 30-60cm from the plane of the transmitting terminal, and was successfully used to supply power to three electric devices (LED lamp, small fan, and mobile phone).

Claims (1)

1. A parallel feedback communication method based on a multi-input multi-output wireless charging system is characterized by comprising the following steps:

step 1: collision-aware parallel decoding scheme

RXs will respond concurrently to TX emulated pings and feed back their necessary data, in the proposed MRC-WPT system, RX will switch the tank circuit to bothThe possible states encode their data, "open (O)" and "closed (S)". When Q RXs are transmitted simultaneously, the collision signal will have 2^QThe combined state is obtained by measuring TX current, performing coordinate coordination among all TXs, identifying RXs combined state, simultaneously estimating mutual inductance coefficient of TXRX/RX-RX, fixing TX voltage to a constant value and taking TX current as a measurement variable;

step 2: observation of phenomena

Clustering phenomenon-the presence of RXs affects the TX current value given that the TX voltage is constant. We performed experiments on our MRC-WPT prototype test bench using 2 TXs and 2 RXs, we fixed the TX input voltage to 8V and switched the RX oscillating circuit to an open state, implemented RX feedback communication, collected and plotted the TX current value, which was significantly affected by the RX on-off state switching.

And step 3: construction of single hop transition graphs

Based on the symbols collected in the n-dimensional space (receiving end switch state representation), we construct a single-hop transition Graph (OFG), in which the vertex set contains 2^QA cluster of symbols and an edge set contains any pair of adjacent edges.

According to a circuit equation, the cluster distribution of the MIMO MRC-WPT system is more regular than that of the RFID communication. When dominant RXs of K ≦ Q is present and its effect on TXs is much greater than the other RXs, 2^QThe clusters tend to form 2^KEach group consisting of 2^Q-KAnd (4) cluster composition. The classifier may not be able to distinguish between clusters that are distributed tightly within a group, resulting in incorrect classification results. As shown in fig. 4(a), each group has 4 groups, each group has 2 clusters, and the classifier has a result of a judgment error.

To improve the accuracy of classification, we use a two-level clustering mechanism, at the first level, we try to locate 2^KWhen there are a sufficient number of symbols, the distribution of symbols in all clusters is normal according to the circuit equation. Thus, the parameter K of the first layer can be adjusted to achieve an even distribution of symbols among groups.

At the second level, we further classify each group,to obtain 2^Q-KAnd (4) clustering. FIG. 4(c) shows the second layer clustering results of the upper left group of FIG. 4 (b). The LDBC algorithm introduced therein is used to locate the center of the cluster in both the first and second layers.

After determining the center of each cluster, we will perform a reliable symbol allocation. A symbol is classified as a cluster if the probability (determined by the distance of the symbol position from the cluster center) of the symbol in the spatial domain is greater than a given threshold. For confusing symbols that may exist in multiple clusters of overlapping spaces, the classification is based on a joint consideration of the probabilities in the space and time domains, where the probability in the time domain is determined by neighboring symbols within the same time window. After symbolic clustering, the next step is to build connections in the graph. Based on the observation that the transition probability between neighboring clusters is significantly higher than the transition probability between non-neighboring clusters, we can identify neighboring clusters from the transition probabilities.

And 4, step 4: potential cluster identification

RFID communication-related research proposes a layer-based cluster identification algorithm that starts with a particular anchor cluster, assuming its combined state is known. This assumption is reasonable in the RFID scenario because the charged tags do not affect the signal received in the IQ domain, and when all tags are charged, clusters representing all "low" states can be identified.

And 5: calibration based on channel estimation

For any given cluster identification candidate, we can obtain the mutual inductance of TX-RX and RX-RX, respectively, from clusters with one or two "closures" RXs. After all the mutual inductances are known, we further evaluate the superiority and inferiority of a given cluster identification by comparing the expected and measured TX currents. The reason for this is that the mutual inductance obtained by the different clusters will be consistent with other clusters in the effective cluster recognition and will conflict with other clusters in the ineffective cluster recognition. Therefore, we will select the best cluster identification by calibration based on channel estimation.

Step 6: decoding

Finally, we can perform decoding based on the best cluster identification, where each RXs combined state of opening and closing is identified. By examining these combined states, the scheme outputs a sequence of "open" and "closed" to represent the signal transmitted by each RX. In collision scenarios, an error label during symbol clustering may increase the error rate of a single RX. The standardized FM0 and Miller predictable state flipping patterns present a possible solution to correct errors in the state sequence before it is input to the decoder.

Images