PhD Thesis Defense · Karadeniz Technical University · Graduate School of Natural and Applied Sciences · 2022
Analysis of Smart Antenna Based Scheduling Algorithms for IoT Networks
Summary
At a Glance
6TiSCH schedules radio resources in time and frequency. When the 2.4 GHz band is crowded and networks are dense, both run short. Smart and directional antennas add a third dimension: space.
A False Beacon lets nodes learn neighbours and their directions before they synchronize.
RSSI-based best-direction tracking keeps the strongest antenna direction for each neighbour.
Two hybrid schedulers: distributed + autonomous, and centralized + autonomous.
1 · Problem Definition · Interactive
The Crowded 2.4 GHz Band
IEEE 802.15.4 offers 16 channels, but Wi-Fi, Bluetooth and other 802.15.4 networks share the band. Toggle Wi-Fi channels:
1.2–1.3 · Problem and Motivation
Constrained Nodes, Shared Cells, Wasted Energy
- Sensor nodes are tiny: e.g. TelosB has a 250 kbps 802.15.4 radio, an 8 MHz MSP430 and 10 kB RAM. They decide with local knowledge only.
- Nodes in hard-to-reach places live as long as their battery: energy decides network lifetime.
- New nodes join through shared cells: every node listens on the same slot and channel, so collisions and interference are frequent.
- Omnidirectional antennas radiate everywhere; current radio duty-cycling ignores antenna direction, and standard MAC protocols do not fit directional antennas.
(1) smart antennas reach further with fewer hops, (2) scheduling wakes radios only when needed, (3) smarter schedules cut delay and keep control traffic healthy.
Using smart antennas to improve 6TiSCH had not been studied: resources stayed two-dimensional.
1.4 · Aim and Contributions
Contributions
Switched-beam and steerable antenna models, configurable from the scenario file (number of antennas, beamwidth).
Direction-aware resource planning, a neighbour table that keeps the best antenna, tests with Minimal Configuration and Orchestra.
False Beacon packet, neighbour-aware beacon direction and an autonomous schedule for control packets.
The direction with the highest received power is stored per neighbour and refined over time.
The known centralized TASA algorithm adapted to antenna directions, then optimized for end-to-end delay.
Cells follow the packet's path to the sink and the reply's path back, minimizing round-trip delay.
2.1 · 6TiSCH Protocol Stack
6TiSCH in Brief
A slotframe repeats; each cell is TX, RX or shared. Channel hopping changes the physical channel every slotframe.
Autonomous: cells from a hash of node IDs (Orchestra). Distributed: neighbours negotiate with 6P (SF0). Centralized: a scheduler with a global view (TASA, AMUS).
A cell becomes (slot, channel, antenna direction).
2.2 · Smart Antennas · Table 1 · Interactive
Smart Antennas and Directional MAC Protocols
Several fixed directional antennas behind an RF switch (e.g. SPIDA with 6 antennas). Choose the antenna that faces the neighbour.
One electronically steered beam: higher gain where it points, protection from other directions, but the right direction must be computed.
| Work | Approach | Limitation |
|---|---|---|
| Schandy et al. | SPIDA, 6 switched antennas, ContikiMAC, RSSI-based antenna choice | asynchronous; several antennas lengthen the slot |
| Voigt et al. | SPIDA in LoRaSim, 4 and 8 dBi | LoRa, not 802.15.4 |
| Varshney et al. | steerable antennas for bulk transfer, 214 kbit/s | no synchronization, single channel |
| Nur et al. | DCD-MAC, DRTS/DCTS | new frame structure, mixes omni and directional |
| Zhang et al. | discovery, allocation and data phases | three-way handshake, costly control traffic |
| Jakllari et al. | PMAC, omni discovery and RTS/CTS | omni control traffic |
| Huang et al. | DBTMA/DA, busy tone | two channels, no multichannel data |
| Korakis et al. | CDR-MAC, RTS in all directions | sweeping all antennas costs time |
3 · Developed Antenna Models
Two Antenna Models in Contiki and Cooja
2 or 3 antennas, each covering 180° or 120°. Broadcasts in shared cells use all antennas; unicast uses only the antenna that faces the destination, learned from the antenna that received its Enhanced Beacon.
Two firmware registers hold direction and beamwidth. Gain follows Eq. (3): G(θ) = 10(1 + cos θ) − 15 dBi. RSS follows the log-normal path-loss model of Eq. (4) with Ptx = 0 dBm, path loss −40 dB at 1 m and exponent 2.5.
3 · Eq. (2)–(4) · Interactive
Antenna Lab: Patterns and Link Budget
0° = pointing at the other node. Threshold from the 50 m omni range in Cooja; Table 6 lists 15–80 m for the smart antenna.
3–6 · Spatial Reuse · Interactive
Spatial Reuse Lab: Same Slot, Same Channel
Illustration, not a thesis result: 25 nodes 40 m apart, one routing tree, a greedy TASA-like scheduler; two links share a slot and channel only if neither receiver hears the other sender above −90 dBm. Colours are channels.
4.1 · Switched Beam with Distributed Scheduling
SF-DA: Direction-Aware Distributed Scheduling
for each slot in slotframe:
if SHARED slot:
if buffer empty: listen on all antennas
on reception: update neighbour(dst, antenna)
else: transmit on all antennas
if TX slot and buffer not empty:
if broadcast: transmit on all antennas
else: transmit on antenna[neighbour(dst)]
if RX slot: listen on the antenna
facing the scheduled neighbour- Cells are triples: (slot, channel, antenna offset).
- Statistics are kept per antenna, not per cell: antennas facing different directions see different channel conditions.
- Parallel transmissions on the same channel become possible when they point away from each other.
- Shared cells still use all antennas, so joining and control traffic keep working.
4.3 · Test Results
Switched Beams: What the Tests Showed
| Setting | Value |
|---|---|
| Topology | 25-node grid, 200 m × 150 m, root in the centre |
| Slot / slotframe | 15 ms / 15 slots |
| Ranges | 50 m transmission, 100 m interference, nodes 40 m apart |
| MAC | backoff 1–7, 5 retransmissions, buffer 5, 127-byte frames |
| Runs | 2 hours, 10 random seeds |
With Minimal Configuration and Orchestra, directional antennas raise packet delivery 2–3×; more antennas, narrower beams, more gain.
Up to 100% delivery improvement when channels are scarce; the effect grows as link quality drops (90%, 80% reception).
With more than 4 channels the gain disappears: directional antennas pay off in dense, interference-limited networks.
4.3, 5.6 · Tables 3–4 · Interactive
Delivery, Delay and Energy against Network Size
4 channels, 100% reception probability, 1 packet/s per node. Table 3: 25, 49, 73, 101 nodes; Table 4 adds 81 nodes and the steerable antenna.
5 · Steerable Antenna with Distributed Scheduling
Why a Narrow Beam Slows Down Joining
Enhanced Beacons are lost to collisions and interference in the shared cell.
A joining node listens on another channel than the one the beacon uses.
A beam pointing the wrong way cannot hear the beacon at all.
False Beacon (5.2). Unsynchronized nodes broadcast a False Beacon instead of an EB. Other unsynchronized nodes keep their radio on, so they hear it easily and store the sender, direction and RSSI in a virtual neighbour list, before joining the network.
Neighbour-aware direction (Alg. 4). In a shared cell a node points to a random neighbour from the list, or with probability 1/(neighbours + 1) to a random direction, to still find newcomers.
on packet from id: (Alg. 3)
n = neighbour(id)
if n is new: add(id, dir, rssi)
else if rssi > n.rssi:
n.dir = dir; n.rssi = rssi
# at most 8 neighbours → Θ(1)5.3 · Eq. (5)–(7) · Interactive
Autonomous Cells for Broadcast Traffic
One steered broadcast cannot reach all neighbours, so each broadcast is copied per neighbour and sent as unicast, in SF0 cells or in autonomous cells hashed from the two IDs.
tTX = ((IDk+1)·7 + (IDo+1)·13) mod 251 · tRX = ((IDo+1)·7 + (IDk+1)·13) mod 251 · f = (IDo+IDk) mod Lfas
Extra cells per node: Ca = 2·Nn·Ls/La, growing with the number of neighbours, not with network size.
5.6 · Test Results
Steerable Antenna: Results
Joining with a narrow beam stays slower than with an omni antenna, but autonomous cells gave the best synchronization of all variants.
Duplicated broadcasts cost energy when links are perfect; with worse links or heavier traffic the steerable antenna wins.
Up to 80 m reach means fewer relays, so poor links hurt less than with 50 m omni links.
6.1 · Steerable Antenna with Centralized Scheduling · Table 5 · Interactive
Direction-Aware TASA (DATASA)
Nodes report parent, neighbours and measured directions to a scheduler behind the border router. The scheduler builds the connectivity, interference (2-hop), conflict, direction and traffic matrices, adds autonomous cells, then repeats per slot:
- Matching: conflict-free links, busiest nodes first.
- Colouring: assign channels; links that point away from each other may share one.
Example with 20 nodes and one channel: 32 slots with omni antennas, 25 with steerable.
6.2–6.3 · Path-Based and Bidirectional Scheduling · Table 7 · Interactive
Scheduling Along the Packet's Path
A node that just received forwards first, so packets skip relay buffers: 10–15% lower delay.
A PCE schedules cells hop by hop to the sink and the reply back; up to 20% less energy than distributed.
Table 7, request + reply delay. Centralized always beats distributed; at 49 nodes the steerable antenna overtakes omni in distributed scheduling.
4–6 · Synthesis
Which Approach, When?
Switched beam
- simple RF switch, no direction search
- lowest energy (narrower beam, less power)
- fits distributed 6TiSCH directly
Steerable
- longer reach, fewer hops
- best delivery in large networks (82.6% at 81 nodes)
- needs discovery help: False Beacon, autonomous cells
Distributed
- scales, adapts locally
- random cells: higher and more variable delay
Centralized
- shortest schedules, lowest delay and energy
- does not scale: tested up to 49 nodes; static, small networks
7 · Conclusion
Conclusion
- Switched beams in 6TiSCH cut collisions in shared cells and improve delivery, energy and delay when channels are scarce.
- Steerable antennas need help to join; a hybrid of distributed and autonomous scheduling and the False Beacon solve synchronization, broadcast and discovery problems.
- A bidirectional, path-based centralized scheduler gives lower delay and energy than distributed scheduling in small, static networks.
- Scheduling in time, frequency and space outperforms time and frequency alone in delivery, energy and delay.
7 · Future Work · Related Publications
Future Work and Publications
- Validate on a real testbed with switched and steerable antennas.
- Evaluate under heavy interference from other 2.4 GHz technologies.
- Machine learning for best-direction search and for cell selection in distributed scheduling.
- Integrate multipath routing with steerable antennas.
Review
Check Questions
- Why does a crowded 2.4 GHz band limit 6TiSCH, and what does a directional antenna add?
- Why do broadcasts in shared cells use all antennas, but unicasts only one?
- Why does a steerable antenna make joining the network slower, and how does the False Beacon help?
- How can two neighbours find a common autonomous cell without exchanging messages?
- When is centralized scheduling the better choice, and when is it not?
Show answers
- Only a few channels stay free of Wi-Fi, so frequency diversity is lost; directional antennas let links that point away from each other share a slot and channel (spatial reuse).
- In a shared cell any neighbour may be listening, so all directions must be covered; a unicast's destination is known from the neighbour table, and one antenna causes less interference.
- The beam must point at the right neighbour at the right time on the right channel; the False Beacon lets unsynchronized nodes learn neighbours and directions in advance, so beacons go in useful directions.
- Both hash the pair of IDs (Eq. 5–7): one node's TX slot equals the other's RX slot, on the same channel offset.
- In small, static networks with scarce channels and strict delay needs; not in large or changing networks, where collecting topology and distributing schedules costs too much.
References
Selected References
- Vilajosana, X., Watteyne, T., Chang, T., Vučinić, M., Duquennoy, S., Thubert, P. (2019). IETF 6TiSCH: A tutorial. IEEE Commun. Surveys Tuts., 22(1), 595–615.
- Palattella, M. R., Accettura, N., Dohler, M., Grieco, L. A., Boggia, G. (2012). Traffic aware scheduling algorithm for reliable low-power multi-hop IEEE 802.15.4e networks. IEEE PIMRC, 327–332.
- Duquennoy, S., Al Nahas, B., Landsiedel, O., Watteyne, T. (2015). Orchestra: Robust mesh networks through autonomously scheduled TSCH. ACM SenSys, 337–350.
- RFC 8180 (2017). Minimal IPv6 over the TSCH Mode of IEEE 802.15.4e (6TiSCH) Configuration. IETF.
- Schandy, J., Olofsson, S., Gammarano, N., Steinfeld, L., Voigt, T. (2021). Improving sensor network performance with directional antennas: A cross-layer optimization. ACM Trans. Sens. Netw., 17(4), 1–21.
- Varshney, A., Mottola, L., Carlsson, M., Voigt, T. (2015). Directional transmissions and receptions for high-throughput bulk forwarding in wireless sensor networks. ACM SenSys, 351–364.
The thesis cites 140 references; see the full text. Charts and diagrams on these slides are redrawn from the thesis.