This document aims to present the general ideas and methodology that may be used in order to design, develop and finally conduct performance tests in order to optimize the metrics of each of the evaluation criteria (i.e. reliability of the send data, end-to-end latency and and power consumption).
[Wireless communication]
Department of Network Engineering Universitat Politecnica de Catalunya, Spain [email protected]
Department of Network Engineering Universitat Politecnica de Catalunya, Spain [email protected]
Mobile and Wireless Internet Unit i2CAT Foundation, Spain [email protected]
Department of Network Engineering Universitat Politecnica de Catalunya, Spain [email protected]
According to EWSN 2016 Dependability challenge specifications, we devise several solution candidates for the high-interference wireless sensor network environments to achieve high energy-efficiency (measured as average energy consumed per successful application layer packet transmission) and low latency. The challenge uses off-the-shelf Maxfor MTM-5000 sensor nodes, which bring some computational and memory constraints.
We take two approaches for this challenge: i) optimization of standardized communication protocol stack through tweaking the parameters of different protocols within stack as we have done in @Betzler14, ii) a cross-layer approach that uses stateless routing as in @Feng12. In the following we detail these two approaches. For the first approach, we employ the protocols devised by Internet Engineering Task Force (IETF) for Internet of Things (IoT) applications.
IETF has developed several solutions for each layer of the OSI protocol stack considering the limitations of the IoT devices could bring. We will be using the stack that is composed of CoAP/UDP/IPv6-RPL/6LoWPAN/IEEE 802.15.4. As an option, the CoCoA algorithm @Betzler15 proposed for congestion control over CoAP is also considered within this stack. The derivation of optimal parameter settings for different layers will be studied for interference-environments. As an example, the RPL protocol parameters investigated are minimum Trickle interval size (Imin), the RPL DIO redundancy (K), and the Parent Switch Threshold (PST) defined by the Objective Function used. These parameters determine how frequently the RPL control messages are transmitted. The clear trade-off of network stability vs. amount of overhead traffic is assessed within this work.
A clear solution for the MAC and PHY layers for interference environment is the Timeslotted Channel Hopping (TSCH) mode of IEEE802.15.4e that has been standardized in 2012. TSCH enables low-power operation through time synchronization and channel hopping for high reliability. As the targeted platform does not include an RF chip with this functionality, one option considered in this work is the implementation of the MAC layer of this standard.
Even though using the inherit capabilities of the well-known IEEE 802.15.4-based stacks such as ZigBee or IETF IoT stack, one can implement a functional solution for the targeted environment, it is clear that there are inefficiency incurred, such as the encapsulation of network, transport and application protocol packets, missing communication of state information between layers, etc. Taking this into account, a potentially simpler and leaner solution for the specific application of the challenge can be developed.
For this, we consider a stateless routing protocol that incorporates the MAC layer functionalities. Regarding the kind of topology and the challenge requirements, a stateless routing is expected to considerably increase the end-to-end performance, while allowing an easier implementation after a short network discovery phase at the beginning. The purpose of this discovery phase is to assign rank numbers (number of hops to reach the sink) to nodes. The basic idea of the method is then the following. When a sender node has some data ready to send, it sends an RTS with its rank number. The available relay nodes with a higher (and possibly equal) rank will send a CTS back, then the data transmission will be carried out. This approach does not require any neighbor information, nor routing table information, allowing an asynchronous system and it is impervious to any broken links or nodes. MAC reliability is enabled through ACK packets.
Another important aspect that must be taken into consideration is how the interference nodes can be avoided. A channel hoping and duty cycling approach will be used for this purpose.
In order to devise and evaluate the solutions involving different network topologies and different strategies, there are several evaluation methodologies that can be used.
An available tool for the two operating systems for the constrained nodes, TinyOS and ContikiOS, is Cooja Simulator. This software allows the user to establish the desired network topology, with different number of devices and surrounding objects, obstacles or interferences. It also allows collecting the performance information such as delays, errors and energy consumption, which are the target to beat within the challenge context.
However the ultimate results of a simulation only match to predictions and estimations based on the possible behavior, actually treating it as much as possible with real capabilities.
Another approach to perform tests and check possible programming issues involves the use of a group of physical nodes in which each of them are separately programmed, including at least one sensing and one sink node and few other intermediate node in order to test the behaviour of the developed algorithms. However, a larger size TelosB testbed is available in our university, consisting on a large grid topology of 6x10 TelosB boards, already used in evaluations of @Betzler14. This testbed can be used for physical environment tests and validation of simulation results.
Due to embedded system capabilities regarding the available boards, two OS options become potentially useful within this context: TinyOS and ContikiOS. A further more detailed analysis would be performed in order to choose which one fits better with the project requirements.
This work was supported in part by the Spanish “Government’s Ministerio de Economia y Competitividad” through project TEC2012-3253, RYC-2013-13029 and FEDER.