| Carl Friedrich Gauß Faculty | Department of Computer Science

Adapting InPhase for large indoor networks

Student(anonymous, Login required)
SupervisorYannic Schröder
ProfessorProf. Dr.-Ing. Lars Wolf
IBR GroupCM (Prof. Wolf)
TypeMaster Thesis


Modern industry and commerce rely on various networked systems to exchange information between different systems and people. If the network is wireless, knowing the position of freely moving devices has multiple useful applications. It can be used to track goods and equipment, or provide location-based services.

In outdoor environments, the Global Navigation Satellite Systems (GNSSs) are available as cheap and established localization systems. Unfortunately, indoor deployments are usually unable to utilize them, as they cannot pick up satellite signals. This issue is solved by indoor positioning systems (IPSs).

InPhase is an IPS developed and researched at the Institut für Betriebssysteme und Rechnerverbund (IBR). It is intended to be used in wireless sensor networks (WSNs). These consist of low-cost,energy efficient devices, usually called sensors nodes, that are deployed sparsely to cover large areas. The core technique behind InPhase is phase-based ranging. Some radio transceivers, like the AT86RF233, can measure the phase angle of radio waves. By exchanging signals at different frequencies, this can be used to calculate the distance between two nodes.

To determine the position of a mobile node, it performs the ranging procedure with multiple anchors, which are nodes that have a fixed, known location. The distance measurements are collected at a central server. There, a position that fits all measurements can be computed. Because measurements are impaired by external factors like radio noise, they are filtered. Currently, InPhase employs a particle filter (PF) to perform those tasks. The accuracy of this method depends on the quality of measurements and the number of available anchors.

InPhase, theoretically, is usable with any WSN whose radios allows for phase-angle measurements. In practice, efficient networking and proper estimation needs additional research.


This thesis aims to improve InPhase’s capabilities in large deployments.

As an exemplary use case, one might picture an industrial factory floor, housing various portable goods and containers. As they are moved around by people and machinery, their position should be traceable.

To facilitate this, the floor can be equipped with a network of InPhase compatible nodes. Anchors are spread throughout the area. Small, battery powered nodes, henceforth called tags, are attached to moving objects. A subset of the anchors acts as coordinators, which have a communication link to the central backend server. The backend software can be queried about the position of any tag. It will then send out instructions to conduct measurements, collect the data and finally estimate the position. The coordinators act as intermediaries between other nodes and the backend. The following functional requirements need to be considered:

  • Low cost: The system should work with minimal resources. It should be reliable withoutadditional sensors.
  • Low energy: Tags, as well as some of the anchors, are battery powered. Energy should thereforebe conserved as much as possible. The nodes need to keep their radio usage to a minimum.
  • Low latency: A tags position should be determined quickly.
  • Simple setup: Localization should only rely on the known anchors. Any manually-defined maps would be unreliable, since the environment is frequently changing. Knowing the ID of a new tag should be enough to localize it, without having to reconfigure the system.

This is somewhat achievable with InPhase, but communication protocols and backend are only implemented naively: Every anchor and tag is predefined and hard-coded on each node. If the position of a tag is requested, all anchors conduct ranging with the tag continuously. The current approach is unsuitable for any large, diverse deployment: It is lacks resource-saving decision-making, and does not adapt to changes.

New tags should automatically be detected and registered. This can be done with some kind ofbroadcast announcement - whether this should be done by the tag or the anchors needs to be determined.

Occasionally, tags may be in range of an anchor which is not directly connected to any coordinator. This may be problematic, because an anchor will also have measurement data that needs to be reach the central back-end. Fortunately, it can be assumed that all the anchors form a largely static network. With some existing protocol, like RPL or parts of the RIME stack, data can be forwarded along the anchor network to a coordinator. Using existing protocols reduces the work-load for this considerably.

Instead of performing the ranging procedure with every anchor, a suitable set of anchors should be selected to find the position with as few ranging attempts as possible. The subset of anchors should be small, to reduce measurement duration and therefore failure probability, while delivering as much information as possible. If the position cannot be estimated after one round of measurements, repeated ranging should be done selectively, possibly with additional anchors. Anchor-selection can be done by examining the connection-quality between the anchors and the tag. Received Signal Strength Indicator (RSSI) and Link Quality Indicator (LQI) values, provided by the AT86RF233 radio transceiver, can be used for this. Currently the PF needs multiple measurement rounds to converge on a position, and may produce a false positive. It is always initialized across the whole area. This may be pruned if all the regions with no radio connectivity are excluded


last changed 2020-11-06, 09:13 by Yannic Schröder