ADR Algorithm Outline

An ADR algorithm consists of two parts: a Link Quality Estimator (LQE) and a de- cision ‘module’.

LQE is the process of estimating the quality of a link. Often this estimation is dis- tilled in a single metric called Link Quality Indicator (LQI). The quality of link is in- fluenced by external interference and internal interference. External interference is the interference caused by the physical environment, such as length of the path between sender and receiver (path loss), other radio sources, and physical objects and the en- vironment itself on that path, causing effects like reflection, attenuation, shadowing,

and multipath fading. This interference varies over time, by changing environmental circumstance (weather, air pressure, people moving). Therefore it is necessary to be able to determine whether the current settings meet the quality requirements, and how much more room for improvement there is, i.e. what the link margin is.

Internal interference is the interference caused by other nodes in the network, by colliding or attenuation with our transmissions. Since these nodes are under our con- trol, the effect of this type of interference can be mitigated. This is one of the respons- ibilities of the MAC layer, to determine which node can send when on what condition. There is a wide-range of strategies, and many algorithms have been published over the years, each with their own trade-offs in terms of complexity, speed, and autonomous control.

To estimate the quality of a link, one could send a large amount of packets and count how many arrive. As this is quite energy inefficient, the challenge is to estim- ate the link quality from as little information as possible, minimising the amount of (additional) packets required.

When assessing the link quality based on received packets, it is important to de- termine whether reception errors come from physical interference, or from other nodes. If a particular link quality is considered to be too low due to physical inter- ference, lower rates and higher transmission powers help in combating the physical interference. However, if the poor performance is caused by a high traffic load, a lower data rate would actually increase congestion, leading to more potential nodes in the collision range. A better strategy in this case would be to, counter-intuitively, increase the data rate, assuming there is enough link margin left. Therefore it is important to determine the source of an error, as changing the data rate may have an opposite effect to the one intended.

Once we have established the quality of a link, a decision module can decide what would be the next step: could there be a better, more energy efficient setting to use,

5.2. Transmission Parameters Selection 95 should we stay with our current setting, or should we actually switch to a lower data

rate, or increased transmit power. As LoRa offers a wide range of possible settings, whereby the difference between settings may be small, trying out every single setting would not be efficient. Therefore, a decision module has to make a ‘smart’ choice on what setting to pick. It can change the transmission power or data rate based on different criteria, like ordering the data rates and picking a faster data rate, or a lower transmission power, if there is enough link margin available.

An ADR algorithm generally has the choice of either only controlling the power, only controlling the rate, or do combination of the two. Each option has its own trade- offs in improving energy efficiency, but potentially also reducing the link reliability. An ADR optimisation algorithm therefore has to strike the balance between energy efficiency and reliability.

Power Control

Using only power control is the easiest option to implement, as it does not require any coordination between nodes and gateway. In a LoRa network, both sender and receiver need to be configured for the same data rate (SF and BW) for a transmis- sion to be successfully received. When using a LoRaWAN compatible gateway, this can be mitigated, as it supports receiving a transmission on any SF as long as BW is 125 kHz. Only changing the transmit power does help in making a node more energy efficient, by ensuring it uses the minimal amount of energy required to reach the gate- way. Power control does not increase the capacity of the network, compared to using different data rates, unless one considers the collision domain. To help a node in op- timising its transmit power, gateways could return the Received Signal Strength (RSS) in ACKs, so nodes can adjust their transmit power based on the desired link margin, although the correlation between RSS and PRR is weak (see section 5.3.3). Gateways can also keep track of the PRR and the RSS of received packets, and instruct nodes

to adjust their power, or let the node decide itself what to do. In this scheme, a node would be fixed to use the lowest data rate that guarantees a good link. Depending on the distance of the node to the gateway, some savings can be made in energy consump- tion, though more savings can be made if we also adjust the data rate.

Rate Control

Only changing the data rate is the most effective way to reduce energy consumption, especially in LoRa, as each step up (in either SF or BW) doubles the data rate, and there- fore halves the airtime and energy consumption. It is harder to implement, however, as it requires coordination between the sender and receiver (i.e. node and gateway). Both parties need to be configured to use the same data rate (i.e. same SF and BW) to be able to receive messages. This can be mitigated by using a receiver with a LoRa baseband chip like the Semtech SX1301 (often used in LoRaWAN gateways), as these chips can receive transmissions on any SF as long as BW is 125 kHz.

Combining of Rate and Power Control

The most gains can be made by combining the data rate and power control, also known as joint power and rate control. Usually, rate control is applied first, and then power control is used if there is enough link margin left.