SIU 2022 · 30th Signal Processing and Communications Applications Conference · IEEE
A New Routing Objective Function for IETF 6TiSCH Protocol
Summary
At a Glance
Every RPL parent switch in 6TiSCH forces a node to give up its cells and negotiate new ones. Two new objective functions widen the switching threshold for nodes that matter most: those with many neighbours (SR1) or many transmit cells (SR2).
≈ values are read from the plots in Figs. 3–5; the paper reports them only graphically.
I · Introduction
RPL Routing on Top of 6TiSCH
- IEEE 802.15.4e TSCH gives industrial IoT reliable, bounded-latency links; 6TiSCH joins TSCH networks to IPv6.
- RPL builds a DODAG, a tree-like graph rooted at the border router. The root starts with a DIO broadcast; nodes answer with DAO messages. DIOs are paced by the Trickle timer.
- Each node computes a rank, its distance from the root. To avoid loops it may only choose parents with a lower rank.
- The objective function (OF) defines how rank is computed and how the preferred parent is chosen. Standard choices: OF0 and MRHOF (minimum rank with hysteresis).
MRHOF usually uses ETX, the expected number of transmissions per packet: good links give low rank.
A candidate must beat the preferred parent by a threshold: 96 rank units (0.75 ETX) in MRHOF implementations, 192 (1.5 ETX) in RFC 6719.
III · Fig. 2 · Interactive
Why a Busy Router Should Not Switch
- Topology after Fig. 2: green root, orange critical router. Cells (timeslot and channel) are allocated per parent and must be re-negotiated after a switch.
- Meanwhile all packets of the subtree wait in node 4's buffer; overflow is dropped.
- With a low, fixed threshold, similar-rank parents cause frequent switches: parent flapping.
II · Related Work
Related Work
RFC 6719 and RFC 6550: rank from link metrics such as ETX or hop count; RPL does not mandate a particular OF.
Better ETX estimates in 6TiSCH lead to paths with fewer transmissions (Fanucchi et al., 2019).
Jin et al. (2020) raised the delivery ratio, but at the cost of higher end-to-end delay and more energy.
Farag et al. (2020) switch parent when a queue backlog factor exceeds a threshold, using a metric of path length, link quality and queue load.
III · Proposed Mechanism · Algorithms 1–2
SR1 and SR2: A Threshold That Grows with Importance
input: parent_1, parent_2
if tx_cells == 0:
cost(preferred) += max_link_metric
else:
if tx_cells > 0 and neighbours > 1 and n < 6:
threshold = n × etx_divisor
else:
threshold = 6 × etx_divisor
if |cost_1 − cost_2| < threshold:
return preferred parent
return lower-cost parent
n = neighbours (SR1) or tx_cells (SR2)The current parent gets a penalty of the maximum link metric, so a node without cells is free to move.
A node with many neighbours likely routes for others; its threshold grows to n ETX.
Cells mirror the traffic through the node; more cells mean a costlier switch.
The threshold stops at 6 ETX, so a switch is never made impossible.
III · Proposed Mechanism · Interactive
Threshold Playground: Keep or Switch?
| OF | Threshold | Cost of preferred | Decision |
|---|
Illustrative costs in ETX (rank / 128). Dot: cost via the preferred parent; band: hysteresis around it; dashed line: candidate. Inside the band the node keeps its parent.
III · Proposed Mechanism · Interactive
Flapping Lab: Two Similar Parents
IV · Experimental Results · Table I
Simulation Setup
| Parameter | Value |
|---|---|
| MAC · simulator | 6TiSCH · Contiki-NG Cooja, UDGM |
| Nodes | 21 (1 server, 20 clients), Exp5438, random placement |
| Reception ratio Rx | 70%, 80%, 90% |
| Objective function | MRHOF, SR1, SR2 |
| Slotframe · timeslot | 21 slots · 15 ms, 6 shared slots |
| Range · packet interval | 50 m · 1 s |
| Duration | 1 h, 5 runs with different seeds, averaged |
Cooja drops the given share of packets at random; lower Rx means lossier links and more reasons to switch.
It is not a realistic reception model, so it was left out.
IV · Figs. 3–5 · Interactive
Results against Link Quality
Values read from the vector plots of Figs. 3–5 (≈); averages of five 1-hour Cooja runs.
IV · Discussion
What the Results Show
Both functions cut parent switches sharply in lossy networks; SR1, driven by neighbour count, is the most stable.
Each avoided switch saves releasing and re-negotiating cells, the window in which buffered packets are lost.
Packets wait less in intermediate buffers, especially when link quality is poor.
At Rx 90% switches are rare anyway and all three functions perform alike.
V · Conclusion and Future Work
Conclusion
- Two objective functions for RPL in 6TiSCH make the parent-switching threshold depend on live node state: neighbour count (SR1) or allocated transmit cells (SR2).
- Nodes switch parents far less often, which raises packet delivery and lowers delay compared with MRHOF, most clearly on low-quality links.
- Future work: tune the threshold with reinforcement learning.
V · Conclusion
Check Questions
- What is an RPL objective function responsible for?
- Why is a parent switch more expensive in 6TiSCH than in plain RPL?
- With 3 neighbours, what threshold does SR1 use, and how does it compare with MRHOF's?
- What happens under SR1 or SR2 when a node has no transmit cells?
- What is the risk of a very large switching threshold?
Show answers
- Computing the rank from routing metrics and choosing the preferred parent.
- Cells are scheduled per parent; switching means releasing and re-negotiating them while packets queue and may be dropped.
- 3 × ETX divisor = 3 ETX (384 rank units), four times MRHOF's 0.75 ETX (96).
- The current parent's cost is increased by the maximum link metric, so the node switches easily.
- The node keeps a parent that has really become bad, so traffic follows a worse path for longer.
References
Selected References
- Winter, T. et al. (2012). RPL: IPv6 routing protocol for low-power and lossy networks. RFC 6550.
- Gnawali, O., Levis, P. (2012). The minimum rank with hysteresis objective function. RFC 6719.
- Levis, P. et al. (2011). The Trickle algorithm. RFC 6206.
- Vallati, C. et al. (2018). Improving network formation in 6TiSCH networks. IEEE Trans. Mobile Comput., 18(1), 98–110.
- Farag, H., Österberg, P., Gidlund, M. (2020). Congestion detection and control for 6TiSCH networks in IIoT applications. IEEE ICC.
- Jin, Y. et al. (2020). Reliable and power confined routing in large and densely deployed 6TiSCH mesh networks. IEEE WCNC.
See the full paper for the complete reference list. Charts are redrawn from the paper's figures.