Most Internet of Things (IoT) systems are based on the wireless sensor network (WSN) due to the reduction of the cable layout cost. However, the battery life of nodes is a key issue when the node is powered by a battery. A Low-Power WSN Protocol with ADR and TP Hybrid Control is proposed in this paper to improve battery life significantly. Besides, techniques including the Sub-1GHz star topology network with Time Division Multiple Access (TDMA), adaptive data rate (ADR), and transmission power control (TPC) are also used. The long-term testing results show that the nodes with the proposed algorithm can balance the communication quality and low power consumption simultaneously. The experimental results also show that the power consumption of the node with the algorithm was reduced by 38.46-54.44% compared with the control group. If using AAA battery with 1200 mAh, the node could run approximately 4.2 years with the proposed hybrid control algorithm with an acquisition period of under 5 s.
Keywords: adaptive data rate (ADR); transmit power control (TPC); time division multiple access (TDMA); wireless sensor network (WSN); power consumption; Internet of Things (IoT)
The wireless sensor network (WSN) is the one of bases in Internet of Things (IoT), and most nodes in the WSN are powered by battery. Extending battery life, saving the maintenance fee, and raising system reliability are the motivations of this paper. In addition to the battery technology improvement and power capacity increase, the low-power technology of the device is also significant. In many applications, IoT device's security and power consumption are significant issues [[
Moreover, wireless transmission is the highest power-consuming process in communication devices. A study pointed out that the power consumption of nodes in the wireless sensing network is mostly concentrated in the process of wireless communication [[
This paper is based on previously published research by this paper's authors, which has discussed the relationship and performance analysis of the transmission power and data rate [[
In summary, this paper provides the following contributions:
- We propose a hybrid control algorithm combined with TPC and ADR that could adapt the environmental interferences.
- Experimental results analysis show that the proposed algorithm achieved energy-saving with stable communication quality.
The architecture of the WSN in IoT application and the selection of communication frequency bands have been discussed in the following studies. The designs for power consumption reduction in the wireless network, such as media access control (MAC), transmission power, and data rate control, have also been also discussed in the following literature.
There are two data processing methods that have been proposed, centralization and distribution data fusion, which each have different benefits. Centralized data fusion processes all the data on a central node, while the nodes in distributed system process their own data [[
Compared with the 2.4 GHz or 5 GHz frequency band, the transmission distance of the Sub-1 GHz wireless communication is farther, so its coverage is wider and its power consumption is lower [[
In wireless communication, transmission power is a major factor for power consumption. A method called TPC minimizes the transmission power as possible when the communication quality can be maintained [[
Figure 1 is the implementation platform of Texas Instruments CC430F6137. The network architecture is a simple star scheme, as shown in Figure 2, and the frequency-shift keying (FSK)/Gaussian frequency-shift keying (GFSK) was selected as the radio modulation method. The network consisted of multiple sensing nodes and a central bridge. The bridge was supplied by grid power and the sensor nodes were supplied by the battery. The placements of the bridge and nodes are shown in Figure 3. The bridge received packet from the nodes, and it was connected by cable to a gateway. The exact position is shown in Section 6.1. The TDMA protocol was proposed in this system, and the data rate and transmission power control algorithms were used to reduce the power consumption of the node to extend the battery life.
Figure 4 is a TDMA diagram proposed in this paper. The concept is to predetermine the communication time slot and turn off the wireless communication function when nodes are idle. The nodes enter the sleep mode for the rest of the time to save power. First, Trigger Time (TTrig.) is the synchronous trigger signal sent by the bridge. Next, Sensing Time (Tsensing) is the time interruption required by the node for sensing, and the time length depends on the processing time required by the installed sensors. Waiting for Require Time (TWRn) is the interval required by the nth node to wait for a bridge command. Response Time (TRes.n) is required to reply to the sensing information packet. Moreover, Delay Time (TDelay) is the slot for packet parsing, wireless communication reception, transmission mode switching, and radio wave calibration for the nodes and bridges. Finally, Acquisition Time (TAcq.) indicates the period from the start of the triggering to the end of the node polling.
The node ID is assigned before the network constructed, and the sequence of the TDMA time slot is dependent on this ID. When a connected sensor node is turned off, the bridge skips the slot after a couple reconnections. The reservation slot is reserved in the last portion of acquisition duration, and it is reserved for the connection of a new sensor node with lowest data rate and highest power. If the node is checked by the bridge, the time slot will be arranged into the dedicated ID slot in acquisition duration.
The crystal oscillator used in this paper had a crystal oscillator error of 10 ppm. Texas Instruments provided the SmartRF Studio tool, which can set and select the appropriate frequency deviation and the receiving channel bandwidth at a specific data rate. The node update period of 5 s was used in this paper, the design goal was 255 nodes. Therefore, the slowest data rate could only be set to 26 kbps. In view of the above factors, the data rate of this paper was divided into nine segments, from fastest (250 kHz) to slowest (50 kHz), to obtain the parameters in Table 1. The paper adopted this method to set the required data rate and the correlation coefficient.
A sensitivity experiment was proposed in this paper, and the detail is shown in Figure 5. Two CC430 RF devices were used in this experiment: The transmitter and the receiver. On the transmitter, adjusting the transmission power is used to change the RSSI, 1000 packets are sent in a fixed data rate, and then average RSSI and PER are calculated on the receiver. In this paper, the corresponding RSSI when the PER was 1% was called sensitivity, and each data rate had a sensitivity. The relationship between data rate, RSSI, and PER is shown in Figure 6. The RSSI closest to one percent PER at each data rate was taken as the receiver sensitivity of the data rate, as shown in Table 2. The above Figure 6 relationship and Table 2 sensitivity table were used to analyze the wireless performance of different data rates.
The transmission powers of 121 segments are provided by the CC430F6137, Sub-1GHz wireless communication chip, while only 41 segments were selected in this paper. Due to the transmission power table provided by the original manufacturer, the actual value did not correspond to the 920 MHz band used in this paper. Therefore, the 121-segment transmission power values were measured at 920 MHz by the Rohde & Schwarz RTO2044 digital oscilloscope with a bandwidth of 4 GHz and a sampling rate of 20 GSa/s, and the more suitable 41 segment settings were chosen in this paper.
Figure 7 is a chart comparing the measurement results of the current consumption corresponding to each transmission power with the original manufacturer. When the transmission power was larger, the difference between the values of the datasheet and the measurements was larger. In the interval where the transmit power was −9 dBm to 6 dBm, regardless of the values of datasheet or the measurements, the consumption variation was suddenly increased. Therefore, the appropriate 41 segment settings from the set value of 121 were selected, as shown in Table 3.
In Equation (
(
The transmission power and data rate hybrid control algorithm was proposed in this paper, and this algorithm was used to balance wireless quality and low power consumption to achieve a PER less than 1% and more power-saving. The control architecture between the bridge and sensor nodes is shown in Figure 9, and the hybrid control algorithm was run in Bridge. In Figure 9, the bridge received the RSSI feedback from sensor nodes' transmission signals and calculated the PER using the packet error interval algorithm. After the hybrid control algorithm was complete, the new transmission power and data rate control command that generated by adaptive algorithm was sent to the sensor nodes from the bridge. The detailed flow of the hybrid control algorithm is shown in Figure 10 and its pseudocode is shown in Figure 11. In this system, input includes the real-time RSSI feedback, PER record, sensitivity table of different data rates, and power consumption table, and output includes the data rate control and transmission power control.
The communication quality target of this paper was set at a PER below 1%. In the algorithm of the packet error interval, the threshold of data rate and transmission power is adjusted by 128 packet durations. If there are no errors in the continuous 128 packets, it means that the PER is less than 1%. The data rate will be increased, or the transmission power will be reduced. However, if there is only one incorrect packet in a 128-packet period, it means that the PER is 1%, and the data rate and transmission power will not be changed.
Then, if two errors have occurred before 128 packets have been completed, the data rate will be reduced or the transmission power will be increased. In this algorithm, when the error interval is short, the larger amplitude of the transmission power is set, because it is necessary to react immediately when an error occurs. Otherwise, when the error interval is long, the lower amplitude of the transmission power is adjusted, because the environment may be stabilized and the error does not occur easily.
Finally, by the error interval method, we determined how much to set the N-grade of data rate and transmission power for the next transmission. After the above N-order adjustment, based on the database of Figure 6 and Figure 8, the lowest power consumption combination of data rate and transmission power was selected as the result of the final adjustment. The method flow is shown in Figure 12.
The experimental location of this paper is the Sixth Hall of Engineering, National Yunlin University of Science and Technology, Taiwan. A total of one bridge and ten nodes were set up in the experiment, as Figure 13 shows.
Every two nodes were placed in the same position. One node had the ADR and TPC control algorithms, and the other node fixed the data rate to the lowest (50 kbps) and the transmission power to the maximum. Through long-term testing, the power consumption and PER of nodes placed at the same position were compared to verify the effect of the algorithm. Nodes 1–5 were the nodes that had the algorithms, and nodes 6–10 comprised the experimental control group. A total of ten nodes were located in five different locations. Note that all nodes ran in the TDMA mode to save a lot of power for the sensing node. However, the power consumption of TDMA is not discussed later.
The experimental data of all nodes is organized as shown in Table 4. Since nodes 6 to 10 were without algorithms, the transmission power was set to a maximum of 10 dBm and the data rate was set to the lowest at 50 kbps. The PERs of nodes 6 to 10 were lower than the PERs of nodes 1 to 5 under the data rate and the transmission power control algorithm. However, the nodes with algorithms maintained a PER of less than 1% except for node 1, and the average current consumption was much lower than that the nodes without algorithms.
Comparing node 1 with node 6, the PER of node 1 was higher than node 6. However, node 1 saved 78.74% more power consumption than node 6 in response packet. In the overall average current consumption, node 1 saved 51.64% more of the energy than node 6.
Node 2 and node 7 were the nearest nodes for the bridge node. In Table 4, the PER of node 2 was 0.6322%, and node 7 was 0.1518%. Although the PER of node 2 was larger than that of node 7, its overall PER was still less than 1%. In the response packet, node 2 saved 73.67% more power consumption than node 7, and in the overall average current consumption, it saved 48.43% more power consumption.
The PER of node 3 was 0.9532%, and node 8 was 0.0912%. In the response packet, node 3 saved 70.49% more power consumption than node 8, and in the overall average current consumption, it saved 46.36% more power consumption.
The experimental result is showed in Figure 14. The PER of node 4 was 0.8429%, and node 9 was 0.3331%. In the overall average current consumption, node 4 was 44.991 uA, and node 9 was 74.281 uA. In the response packet, node 4 saved 59.59% more power consumption than node 9, and in the overall average current consumption, it saved 39.43% more power consumption.
The PER of node 5 was 0.4011%, and node 10 was 0.3376%. In the overall average current consumption, node 5 was 32.516 uA, and node 10 was 74.281 uA. In the response packet, node 5 saved 85.83% more power consumption than node 10, and in the overall average current consumption, it saved 56.23% more power consumption.
Obviously, the RSSI values of nodes with algorithms are close to the sensitivity values corresponding to the current data rate from the experimental results. If the RSSI value is lower than the sensitivity value, the probability of packet error will increase. Moreover, since the position of the nodes is affected by the people in the office and the class, the RSSI value of each node floats dramatically during the daytime and the probability of packet error is high. However, at night, the RSSI value is so stable that the probability of packet error is low. If the node is supplied by AAA battery with 1200 mAh, the execution duration could approach about 4.21 years with the proposed algorithm. The nodes without the proposed algorithm could run only for 1.8 years.
The data of Table 4 is drawn with a bar graph of the nodes' average current consumption as Figure 15. It is clear the response packet of the nodes with the algorithm saved the most significant energy in the TX Mode, and the ranking of the power saving was sequentially ranked as nodes 5, 1, 2, 3, and 4. The reason for the difference in the amount of power consumed by each node was assumed to be the positional relationship.
For justification of the proposed algorithm, the experimental results for long-term testing are presented in Figure 16 and Table 5. Figure 16 shows the battery voltage variation in node 4 and node 9, which ran with and without control algorithm for 69 days of execution. The battery voltage of the node 9 in Table 5 was obviously lower than that of node 4, and the results also verify the proposed algorithm workable.
A wireless sensor network based on Sub1G-Hz and star topology was constructed in this paper, and the TDMA wireless communication protocol and the transmission power and data rate control algorithm were proposed to reduce power consumption on sensing nodes usefully. According to the dynamic environment, sensing nodes with hybrid control algorithms automatically adapt the transmission power and data rate to achieve good communication quality and low power consumption simultaneously. The above algorithms will increase the performance and reduce power consumption on wireless communication. Then, communication devices could be operated at very low power consumption when using wireless communication.
The experimental results show that PER states of nodes can effectively be controlled near the target value, 1%, which can prove the good reliability of communication. In addition, because all nodes run in the TDMA architecture's wireless protocol, TDMA can enable a wireless transmission in a low duty cycle. The average current consumption of the node without the hybrid control algorithm was calculated as 74.281 uA, and the power consumptions of algorithm nodes were different and depended on the positions of the nodes. According to the experimental results, when the power consumption of the response packet in the transmission mode was compared, the power consumption saved up to 85.83%. The overall consumption saved up to 56.23% of the power consumption, which indicates that the algorithms proposed in this paper actually have an energy-saving effect for wireless communication. If the node is powered by AAA battery with 1200 mAh, the node could run approximately 4.21 years with proposed algorithm. The other TDMA is discussed in [[
In summary, the proposed hybrid control algorithm is complex, and the payload and node number in the WSN are also limited. However, the system architecture and control algorithm proposed in this paper could lower several important things such as the power consumption, system complexity, maintenance fee, etc.
Graph: Figure 1 CC430F6137 development board.
Graph: Figure 2 Star network architecture.
Graph: Figure 3 (a) The placement of the bridge; (b) the placement of node 2 and node 7.
Graph: Figure 4 Diagram of TDMA architecture.
Graph: Figure 5 Experimental architecture of receiver sensitivity.
Graph: Figure 6 The relation between RSSI, data rate, and PER.
Graph: Figure 7 Comparison of the datasheet and measured values in the 121 segment transmission power.
Graph: Figure 8 Relation of data rate, transmission power, and power consumption.
Graph: Figure 9 Control architecture between the bridge and sensor nodes.
Graph: Figure 10 Flowchart of error interval algorithm controlling data rate and transmission power.
Graph: Figure 11 The pseudocode of control algorithm.
Graph: Figure 12 Flowchart for selecting the best energy efficient combination of data rate and transmit power.
Graph: Figure 13 Experimental bridge and nodes placement map.
Graph: Figure 14 Experimental results of (a) node 4 and (b) node 9.
Graph: Figure 15 Nodes overall average current consumption.
Graph: Figure 16 The battery voltage of (a) node 4 and (b) node 9 within 69 days.
Table 1 Parameters of different data rates.
Data Rate Frequency Deviation RX BW 250 kbps 126.953125 kHz 541.666667 kHz 225 kbps 114.257812 kHz 464.285714 kHz 200 kbps 101.562500 kHz 406.250000 kHz 175 kbps 88.867188 kHz 406.250000 kHz 150 kbps 76.171875 kHz 325.000000 kHz 125 kbps 63.476562 kHz 270.833333 kHz 100 kbps 50.781250 kHz 232.142857 kHz 75 kbps 38.085938 kHz 162.500000 kHz 50 kbps 25.390625 kHz 116.071429 kHz
Table 2 RSSI when the PER was 1% at different data rates.
Data Rate Sensitivity 50 kbps −96.93 dBm 75 kbps −95.22 dBm 100 kbps −94.36 dBm 125 kbps −93.69 dBm 150 kbps −93.12 dBm 175 kbps −91.96 dBm 200 kbps −91.53 dBm 225 kbps −90.25 dBm 250 kbps −90.06 dBm
Table 3 Current consumptions in 41-segment transmission power.
Transmit Power (dBm) Current (mA) Transmit Power (dBm) Current (mA) Transmit Power (dBm) Current (mA) 10.062 37.739 −1.3157 16.637 −15.688 13.005 9.3152 35.763 −2.0975 16.144 −17.207 12.791 8.2542 33.223 −2.9932 15.633 −18.484 12.612 7.1839 31.071 −4.0892 15.027 −19.539 12.509 6.8546 30.347 −4.7187 14.729 −20.784 12.395 6.2026 29.153 −5.5103 14.42 −21.604 12.356 5.5377 28.042 −6.8004 14.012 −22.212 12.299 4.3618 20.562 −7.5317 13.913 −23.3 12.246 3.6703 20.005 −8.5849 14.893 −24.073 12.173 3.0454 19.349 −9.7606 14.474 −25.125 12.116 2.2062 18.613 −11.155 14.077 −26.202 12.059 0.78139 17.556 −12.904 13.65 −27.653 12.021 −0.14598 17.583 −13.856 13.325 −28.899 11.941 −0.98536 16.942 −14.407 13.219
Table 4 Node experimental results.
Packet Error Number PER (%) Response (TX) Average Current Consumption Overall Average Current Consumption Node1 7989 1.4035 8.4740 uA 35.924 uA Node6 213 0.0374 39.852 uA 74.281 uA Node2 3599 0.6322 10.493 uA 38.310 uA Node7 864 0.1518 39.852 uA 74.281 uA Node3 5426 0.9532 11.760 uA 39.845 uA Node8 519 0.0912 39.852 uA 74.281 uA Node4 4798 0.8429 16.106 uA 44.991 uA Node9 1896 0.3331 39.852 uA 74.281 uA Node5 2283 0.4011 5.6440 uA 32.516 uA Node10 1922 0.3376 39.852 uA 74.281 uA
Table 5 The battery voltage of each node at the 69th day.
Experimental Group Node 1 Node 2 Node 3 Node 4 Node 5 Battery voltage 2.659 V 2.65 V 2.66 V 2.652 V 2.669 V Control group Node 6 Node 7 Node 8 Node 9 Node 10 Battery voltage 2.618 V 2.614 V 2.617 V 2.606 V 2.624 V
All authors contributed ideas, discussed the results. C.-W.H. and H.-J.Z. designed the experiment, and performed most of the analysis. H.-J.Z. compiled the data. C.-W.H. and W.-T.H. wrote most of the main text. Y.-D.Z. supported the important revision of the article. All authors have read and agreed to the published version of the manuscript.
This research was funded by the Ministry of Science and Technology, ROC, grant number under contract No. MOST 109-2221-E-224-023-, 108-2221-E-224-045- and 108-2218-E-150-004-.
To the best of our knowledge, the authors have no conflict of interest, financial or otherwise.
By Chung-Wen Hung; Hao-Jun Zhang; Wen-Ting Hsu and Yi-Da Zhuang
Reported by Author; Author; Author; Author