PaperlayerView exact revision on arXiv

Source-linked AI summary

FREE -- Fine-grained Scheduling for Reliable and Energy Efficient Data Collection in LoRaWAN

Khaled Q. Abdelfadeel, Dimitrios Zorbas, Victor Cionca, Dirk Pesch

arXiv:1812.05744v2cs.NI

TL;DR

LoRaWAN’s Aloha-based access creates collision-related reliability and scalability challenges for large-scale data collection. FREE buffers data for synchronized, scheduled bulk transmissions and allocates transmission resources. It reports more than five-fold higher device lifetime, passing ten years, and >99% data delivery.

  • Problem

    Aloha-based LoRaWAN has high collision probability that limits reliability and scalability, while gateway duty-cycle restrictions also constrain acknowledged traffic.

    Source: §I · p. 1

  • Method

    FREE buffers end-device data and uses synchronized schedules that allocate spreading factors, channels, and transmission powers for bulk collection.

    Source: §I · p. 1, A. Proposed Solution - FREE · p. 2

  • Results

    More than five-fold increase in device lifetime, passing ten years, and a >99% data delivery ratio.

    Source: §I · p. 1

  • Takeaways & Limitations

    FREE is intended for delay-tolerant LoRaWAN data collection, including monitoring and remote-sensing applications.

    Source: §I · p. 1

Abstract

from arXiv · show

LoRaWAN promises to provide wide-area network access to low-cost devices that can operate for up to 10 years on a single 1000 mAh battery. This makes LoRaWAN particularly suited to data collection applications (e.g. monitoring applications), where device lifetime is a key performance metric. However, when supporting a large number of devices, LoRaWAN suffers from a scalability issue due to the high collision probability of its Aloha-based MAC layer. The performance worsens further when using acknowledged transmissions due to the duty cycle restriction at the gateway. For this, we propose FREE, a fine-grained scheduling scheme for reliable and energy-efficient data collection in LoRaWAN. FREE takes advantage of applications that do not have hard delay requirements on data delivery by supporting synchronized bulk data transmission. This means data is buffered for transmission in scheduled time slots instead of transmitted straight away. FREE allocates spreading factors, transmission powers, frequency channels, time slots, and schedules slots in frames for LoRaWAN end-devices. As a result, FREE overcomes the scalability problem of LoRaWAN by eliminating collisions and grouping acknowledgments. We evaluate the performance of FREE versus different legacy LoRaWAN configurations. The numerical results show that FREE scales well and achieves almost 100% data delivery and the device lifetime is estimated to over 10 years independent of traffic type and network size. Comparing to poor scalability, low data delivery and device lifetime of fewer than 2 years for acknowledged data traffic in the standard LoRaWAN configurations.

I. INTRODUCTION

LoRaWAN’s Aloha-based access limits reliable data collection at scale, and acknowledgments can further strain gateways under duty-cycle restrictions. FREE instead buffers data and schedules synchronized bulk transmissions to improve delivery and device lifetime.

I. INTRODUCTION

  • LoRaWAN’s Aloha-based MAC has high collision probability, causing packet loss that limits network reliability and scalability.Acknowledged traffic can add retransmissions while gateway duty-cycle restrictions limit acknowledgments, increasing collisions and energy consumption without improving delivery.Source: §I · p. 1
  • FREE reports a more than five-fold increase in device lifetime, passing ten years, and a >99% data delivery ratio.The paper attributes these improvements to scheduled transmissions that eliminate collisions and longer packets that reduce MAC-header overhead.Source: §I · p. 1
  • FREE buffers end-device data for scheduled bulk transmission instead of sending it as soon as it is generated.This approach is intended for delay-tolerant LoRaWAN data collection applications.Source: §I · p. 1
  • FREE replaces random access with synchronized scheduling, a challenge shaped by duty-cycle limits, concurrent gateway reception, and downlinks following uplinks.The gateway schedule allocates spreading factors, channels, and transmission powers, and acknowledged traffic uses compressed group acknowledgments.Source: §I · p. 2

A. Proposed Solution - FREE

  • Devices with the same spreading factor transmit sequentially, while different spreading factors can transmit simultaneously in parallel frames.Each frame ends with a downlink slot; multiple channels may also be assigned to a spreading factor.Source: Figure 1 · p. 2, A. Proposed Solution - FREE · p. 3

B. Joining and Synchronization Phase

FREE uses two synchronization stages to collect device requests, assign transmission parameters, and disseminate a shared schedule and time reference. Devices then send buffered data in assigned slots, with grouped acknowledgments for confirmable traffic.

B. Joining and Synchronization Phase

  • The joining and synchronization phase first collects device requests and provides transmission parameters with 1-second synchronization, then broadcasts a schedule with synchronization under 1 millisecond.Both stages’ duration and channels are assumed to be shared among devices in advance.Source: B. Joining and Synchronization Phase · p. 3

1) First Stage: Joining the Data Collection:

  • A device’s join request reports buffered data and delay elasticity; the gateway responds with its data rate, power, channels, slot, and timing information.Requests use LoRaWAN’s Aloha access, while the gateway sends the acceptance on the request channel or the extra downlink channel, subject to duty cycle.Source: 1) First Stage: Joining the Data… · p. 3

2) Second Stage: Disseminating Frame Structures:

  • The gateway broadcasts frame settings that give packet lengths, guard times, slots per spreading factor, and timing for the collection round and next synchronization phase.Devices use the packet length and guard time to calculate slot length.Source: 2) Second Stage: Disseminating Frame… · pp. 3–4

C. Data Collection Rounds

  • Devices wake for their assigned slot, transmit, and sleep; for confirmable traffic, the gateway acknowledges received packets together in a bitmap.A device may retransmit an unacknowledged packet up to eight times before dropping it.Source: C. Data Collection Rounds · p. 4
  • Collection periodicity is set by application delay elasticity and must be no longer than the lowest delay requirement.For example, applications with 12-hour and 24-hour delay elasticities require collection every 12 hours or less.Source: C. Data Collection Rounds · p. 4

III. FINE-TUNING A FREE SCHEDULE

FREE’s schedule balances collection time and energy while accounting for LoRa-specific constraints. Its packet-length analysis weighs retransmission costs against MAC-header overhead and finds that longer packets tend to minimize energy.

III. FINE-TUNING A FREE SCHEDULE

  • FREE scheduling must balance collection speed and energy consumption while accounting for technology-dependent factors such as duty cycle and spreading factors.The paper presents these trade-offs alongside allocation algorithms for transmission parameters.Source: §III · p. 4

A. Packet Lengths

  • Longer packets increase error and retransmission risk, while shorter packets increase MAC-header overhead and the number of transmissions needed.The analysis computes packet length to minimize total energy for transmitting buffered data.Source: A. Packet Lengths · p. 4
  • For all spreading factors, the energy-minimizing packet lengths tend to be large because MAC-header overhead outweighs packet-error effects.The evaluation examines packet lengths from 20B to 255B for 1.5KB of buffered data.Source: A. Packet Lengths · p. 5, Figure 3 · p. 5

B. Spreading Factor and Slot Number

Spreading-factor allocation must balance energy use against data-collection time while respecting each device’s link constraints. FREE assigns a spreading factor and then a slot using the selected objective.

B. Spreading Factor and Slot Number

  • Higher spreading factors increase transmission time and energy use, while the allocation also determines frame sizes and overall collection time.The section frames minimizing energy and collection time as competing objectives.Source: B. Spreading Factor and Slot Number · p. 5
  • The allocation algorithm uses α = 1 to minimize collection time and α = 0 to minimize energy consumption.The gateway runs the algorithm after receiving a join request, using its RSSI and buffered-data amount.Source: B. Spreading Factor and Slot Number · p. 5
  • FREE selects a spreading factor by evaluating allowed factors against the chosen objective, then assigns the device the next available slot in that factor’s frame.The gateway uses the join request’s buffered-data amount and RSSI; factors below the minimum supported by receiver sensitivity are excluded.Source: B. Spreading Factor and Slot Number · p. 5

1) Minimum Energy Consumption (α = 0):

FREE offers distinct spreading-factor objectives: reduce energy consumption or balance frame lengths to reduce collection time. The latter can favor a higher spreading factor than the energy objective would choose.

1) Minimum Energy Consumption (α = 0):

  • The minimum-energy cost estimates transmission energy for a spreading factor using average transceiver current and voltage.The passage describes the equation’s inputs but does not provide a readable cost-function expression.Source: 1) Minimum Energy Consumption (α = 0) · p. 5

2) Minimum Collection Time (α = 1):

  • Collection-time minimization balances frame lengths across spreading factors, sometimes assigning a device a higher factor than the energy objective would favor.Devices sharing a spreading factor occupy consecutive slots, and those frame lengths determine the collection schedule.Source: 2) Minimum Collection Time (α = 1) · p. 6
  • A frame’s minimum length is constrained by packet length and duty cycle, allowing a device to reuse its slot number without violating duty-cycle limits.The stated minimum slot count is 1/d when slot length equals packet transmission time Tf.Source: 2) Minimum Collection Time (α = 1) · p. 6
  • The collection-time cost estimates a device’s contribution from the number already assigned to its spreading factor and the available channels.Its inputs include Xf, the number of already assigned devices, and Mf, the number of channels; additional channels can require extra slots.Source: 2) Minimum Collection Time (α = 1) · p. 6

3) Allocation Optimality:

FREE uses online spreading-factor allocation and channel and power assignments to support concurrent transmissions while limiting interference. Its channel plan reflects gateway concurrency and spreading-factor interference constraints.

3) Allocation Optimality:

  • FREE’s online greedy allocation uses information available so far and yields a schedule comparable to the offline allocation, whose collection time is shorter but improves only slightly.The online approach does not require advance global network state and is independent of network topology and application type.Source: 3) Allocation Optimality · p. 6

C. Transmission Power and Channels

  • Because spreading factors are not fully orthogonal, overlapping same-channel signals decode together only when their signal-strength difference is below the corresponding interference threshold.Otherwise, only the strongest signal can be detected; isolation can use power control or frequency assignment.Source: C. Transmission Power and Channels · p. 6
  • FREE assigns one channel to spreading factors 7–10 and two channels to 11–12, keeping the gateway’s maximum simultaneous transmissions at eight.The design assumes at most three uplink channels and eight concurrent transmissions; the extra channel for factors 11 and 12 is intended to reduce their collection rounds by about 50%.Source: C. Transmission Power and Channels · p. 7
  • FREE isolates spreading factor 7 on a separate channel and lowers the transmit power of factors 8 and 9 by 1 dBm to limit cross-factor collisions.The scheme relies more on channel assignment than frequent power control, and its allocation technique is reported to almost eliminate cross collisions.Source: C. Transmission Power and Channels · pp. 6–7

D. Packet Lengths, Guard Periods, and Slots per Frame

FREE determines frame structure after collecting device requests, once the gateway knows the participating devices. It derives slot lengths and frame sizes from packet, guard-time, and duty-cycle requirements.

D. Packet Lengths, Guard Periods, and Slots per Frame

  • After the first stage reveals all participating devices, the gateway computes and broadcasts the frame structure in FSettings messages.The frame structure includes slot lengths and the number of slots per spreading-factor frame.Source: D. Packet Lengths, Guard Periods, and… · p. 7
  • A slot’s length is packet transmission time plus twice the guard time, with packet length based on the largest buffered-data size for that spreading factor.The guard period accommodates the maximum estimated clock-skew accumulation from the previous collection round.Source: D. Packet Lengths, Guard Periods, and… · p. 7
  • The number of slots per frame is raised to the minimum required by the duty cycle when the computed count is too small.Once devices receive FSettings, they know the slot length and slot count for their frame.Source: D. Packet Lengths, Guard Periods, and… · p. 7

IV. PERFORMANCE EVALUATION

The evaluation compares FREE with legacy and delayed LoRaWAN under confirmable and unconfirmable traffic. It uses repeated simulations and measures energy, collection time, device lifetime, and data delivery.

IV. PERFORMANCE EVALUATION

  • LoRaFREE uses Simpy and a measured-environment log-distance path-loss model, adding packet errors, imperfect spreading-factor orthogonality, fading, duty-cycle limits, and reception energy.The simulator also supports bidirectional communication and retransmissions for confirmable uplinks.Source: §IV · pp. 7–8
  • FREE with α = 0 and α = 1 is compared with Legacy and Delayed LoRaWAN for both confirmable and unconfirmable traffic.Applications generate packets according to a Poisson distribution; the scenarios differ in when and how packets are transmitted.Source: §IV · p. 8
  • Each simulation uses a fixed collection period and buffered goal data, and reports the mean and standard deviation across 10 runs with different random seeds.A run ends after all devices transmit their goal-buffer data.Source: §IV · p. 8
  • The evaluation measures network energy and collection time, estimated device lifetime, and network data delivery ratio.For FREE, it also reports spreading-factor distributions, guard times, frame lengths, air time efficiency, transmissions, collisions, and lost packets.Source: §IV · p. 8
  • Air time efficiency is the ratio of ideal to actual data collection time, reflecting overhead from guard times and greedy online spreading-factor allocation.The ideal duration assumes no guard periods or idle time, equal-length frames, and eight concurrent transmissions.Source: §IV · p. 8

A. Joining and Synchronization Phase Study

FREE’s synchronized transmissions deliver a data delivery ratio close to 1 across network sizes in the unconfirmable-traffic study. The evaluation also shows joining and synchronization overhead, along with a trade-off between energy use and collection time across FREE settings.

A. Joining and Synchronization Phase Study

  • In a 2,000-device network, each device transmits an average of 33 join-requests; roughly 17 collide and nearly 14 reach the gateway.The gateway cannot send join-accept messages in both receive windows because of its duty-cycle limit.Source: A. Joining and Synchronization Phase… · p. 9

B. Unconfirmable Traffic Type Study

  • Legacy LoRaWAN devices are estimated to survive about 6 years, compared with 10-plus years for the other approaches, including FREE and Delayed LoRaWAN.FREE consumes less energy than Legacy LoRaWAN but more than Delayed LoRaWAN, including the joining and synchronization overhead.Source: B. Unconfirmable Traffic Type Study · p. 9
  • FREE with α = 0 minimizes energy consumption at the expense of collection time, while α = 1 has the reverse trade-off.The settings use different spreading-factor distributions; α = 1 shifts some low-spreading-factor devices to higher factors to balance frame lengths and reduce collection time.Source: B. Unconfirmable Traffic Type Study · p. 9
  • FREE’s synchronized communications prevent most collisions and achieve a data delivery ratio of almost 1 regardless of network size.Delayed LoRaWAN’s delivery ratio degrades as network size grows, with collisions and channel fading contributing to the decline.Source: Figure 5 · p. 10, B. Unconfirmable Traffic Type Study · p. 11
  • Higher spreading factors require longer guard periods because their data-collection rounds take longer, and their frame lengths are mostly longer than those at lower factors.More balanced frame lengths with α = 1 correspond to higher air time efficiency than with α = 0.Source: B. Unconfirmable Traffic Type Study · p. 12

C. Confirmable Traffic Type Study

For confirmable traffic, FREE sustains long device lifetimes and delivery as network size grows, unlike Legacy and Delayed LoRaWAN. Collection periodicity trades lifetime against collection time, with a 12-hour interval and roughly 2.5 KB buffered data presented as a balance.

C. Confirmable Traffic Type Study

  • FREE devices survive 10 plus years across evaluated network sizes, while Legacy and Delayed LoRaWAN fall below 2 and about 5 years, respectively, at 2,000 devices.Legacy and Delayed incur 7.4× and 6.4× as many transmissions as unconfirmable traffic at 1,000 devices; FREE’s transmission count increases little.Source: C. Confirmable Traffic Type Study · p. 12
  • FREE’s collection-time increase for confirmable traffic is small versus unconfirmable traffic, while Delayed LoRaWAN’s increase worsens with network size because of retransmissions.FREE uses the same spreading-factor distribution in both traffic cases; at α = 1, collection time is lower and energy consumption higher than at α = 0.Source: C. Confirmable Traffic Type Study · p. 12
  • Legacy and Delayed LoRaWAN drop below 20% DDR at 300 devices, with gateway duty-cycle limits—not collisions—the main cause.Uplink and downlink fading also reduce DDR, but have less impact than the duty-cycle limitation.Source: C. Confirmable Traffic Type Study · p. 12
  • FREE’s synchronized transmissions experience almost no collisions, and acknowledgments for a frame are grouped into one message sent in its last slot.The frame length is designed so the acknowledgment slot obeys the channel duty cycle.Source: C. Confirmable Traffic Type Study · p. 12

D. Collection Periodicity and Delay Elasticity Study

  • For unconfirmable traffic, FREE is more energy efficient than Legacy LoRaWAN through hourly collection, while Delayed LoRaWAN is fastest and most energy efficient at the expense of DDR.The hourly boundary and trade-off are reported for the studied periodicities.Source: D. Collection Periodicity and Delay… · p. 12
  • For confirmable traffic, FREE surpasses Legacy and Delayed LoRaWAN in lifetime and collection time without sacrificing DDR across the presented periodicities.Fig. 7 presents results for variable collection periodicity and both traffic types.Source: D. Collection Periodicity and Delay… · p. 13, Figure 7 · p. 13
  • Longer collection periodicity reduces joining and synchronization overhead but increases collection time; the study identifies 12 hours and roughly 2.5 KB buffered data as a balance.The trade-off applies to both traffic types.Source: D. Collection Periodicity and Delay… · p. 13

V. RELATED WORK

Prior LoRaWAN scheduling approaches retain collision risks or make assumptions that limit their scope, while related bulk-transmission work addresses other wireless technologies. The paper concludes that FREE handles LoRaWAN scheduling and acknowledgments with high delivery and long estimated lifetime, while identifying real-network validation as future work.

V. RELATED WORK

  • Earlier LoRaWAN bulk-transmission work assumes known global state and fully orthogonal spreading factors, and uses offline allocation; its lower energy comes with narrower scope and unrealistic gateway signaling demands.FREE instead uses online allocation and addresses synchronization and acknowledged transmissions.Source: §V · pp. 13–14
  • Bulk-transmission protocols for 802.11, 802.15.4, and RFID,, do not address LoRaWAN’s duty-cycle limits and non-orthogonal spreading factors.The paper focuses on reliable, energy-efficient bulk collection over LoRaWAN.Source: §V · p. 14
  • Earlier coordinated LoRaWAN schemes, leave collision risks through Aloha access or shared slots, and their evaluations address unconfirmable traffic only.The schedule-signaling duty-cycle cost is also a scalability concern for the Bloom-filter approach.Source: §V · p. 14
  • The drone-gateway proposal does not buffer bulk data and assumes perfect synchronization, orthogonal spreading factors, and no duty-cycle constraint.The cited work studies unconfirmable transmissions only.Source: §V · p. 14

VI. CONCLUSIONS AND FUTURE WORK

  • FREE’s simulations show almost 100% delivery and over 10 years of battery lifetime across transmission types and network sizes, compared with fewer than 2 years for standard LoRaWAN confirmable traffic.The conclusion attributes FREE’s scheduling to collision-free concurrent transmissions across spreading factors and grouped acknowledgments.Source: §VI · p. 14
  • Real-LoRaWAN implementation and validation, multiple-gateway coordination, and energy fairness are identified as future work.The conclusion also notes that listen-before-talk could expedite collection but might increase device energy use.Source: §VI · p. 14

APPENDIX A

The appendix describes join-request, join-accept, and FSettings message structures used in joining and synchronization, including fields for gateway control of device transmission parameters.

APPENDIX A

  • Fig. 8 depicts the join-request, join-accept, and FSettings message structures used in the joining and synchronization phase.The first three join-request fields retain the meanings used in the standard LoRaWAN message.Source: Appendix A · p. 14, Figure 8 · p. 15
  • The first four join-accept fields retain their standard LoRaWAN meanings, while the added DCSetting field lets the gateway control each device’s transmission parameters.DCSetting contains DataRate, TxPower, ChMask, SlotFrame, and Secondstage subfields.Source: Appendix A · p. 15
  • FSettings encodes PcktSizes, Guards, and FrameLends with successive bytes corresponding to spreading factors 7, 8, and onward.Source: Appendix A · p. 15
Loading 1812.05744v2…