2.1. Consumers

The key requirement for consumer positioning devices is the minimization of cost. This is achieved through large production runs, enabling research and development costs to be shared across many millions of individual devices. Minimization of size, weight and power consumption is also important with the latter achieved by powering up subsystems only as and when required. Performance requirements are secondary and are typically qualitative rather than quantitative. Thus, the position accuracy must be sufficient to enable map-aided route guidance, tracking of people and pets and various location-based services to function effectively. The reliability must be sufficient to avoid unduly annoying the user. Consumers generally accept that new technology does not always work as intended. However, they expect reliability to improve as a technology matures.

Consumers' performance expectations for positioning technology will increase with time. They will increasingly require a seamless position solution, available indoors and outdoors, and when stationary, walking, in a car, on a train, plane or ferry, or playing sport (El-Sheimy and Goodall, Reference El-Sheimy and Goodall2011). This could be achieved in two ways. The first is through the development of a multimodal navigation and positioning system, embedded within a consumer mobile device, which can handle a wide range of environmental and behavioural contexts. The second approach is to achieve seamless positioning collaboratively through the interaction between mobile devices, vehicle positioning systems and infrastructure-based tracking systems. Both approaches require some degree of cooperation between the suppliers of the various subsystems.

2.2. Professionals

For professional users, it is essential that navigation and positioning equipment meets a clearly defined set of performance requirements. For air and sea navigation, failure to meet performance standards compromises safety, whereas for surveying, a failure results in work having to be repeated, increasing costs and compromising schedules. Integrity is thus important. For most professional applications, size, weight and power requirements, though important, are less critical than for consumer applications. The main exception is the tracking of assets and people, for which consumer-grade equipment is typically used.

User equipment costs of several thousand dollars can be justified when compared with the costs of insuring an airliner or ship or the savings in staff time that better equipment can bring. With user bases measured in the tens of thousands rather than tens of millions, research and development accounts for a large part of the equipment cost. Therefore by sharing designs across a wider range of application domains, there is substantial scope to reduce costs.

Different applications have different performance requirements. For air navigation, reliability (expressed in terms of integrity, continuity and availability) is most critical (Rife and Pullen, Reference Rife, Pullen, Gleason and Gebre-Egziabher2009), whereas for surveying, accuracy is more important (Rizos and Grejner-Brzezinska, Reference Rizos, Grejner-Brzezinska, Gleason and Gebre-Egziabher2009). This has led to very different design philosophies that are difficult to reconcile. A modular approach to design would enable common hardware components to be combined with software modules tailored to meet the relevant performance requirements. This approach has already been adopted for GNSS surveying equipment and could potentially be deployed more widely. Some hardware and software components could also be shared with consumer devices.

2.3. Researchers

Researchers comprise both those who use navigation and positioning equipment as a measurement tool and those working to advance navigation and positioning technology itself. Many of the former group are already well served by professional-grade user equipment. Tracking devices are available to support a range of research and development needs while GNSS surveying equipment is used for a wide variety of scientific applications, facilitated by the adoption of a standard format for recording measurement data, Receiver Independent Exchange (RINEX). This discussion will therefore focus on those researchers whose needs are not met by existing commercially available technology. Accuracy and update rate are typically more important than reliability; experiments can usually be designed to circumvent the limitations of the positioning techniques, while faulty data can often be discarded when the results are analysed. The importance of size, weight, power consumption and cost vary with the application.

Researchers using navigation equipment for new applications may need to adapt the technology to meet their requirements. For example, studying the locomotion of wild animals requires the measurement of high dynamics by small, lightweight sensors with low power consumption (Wilson et al., Reference Wilson, Lowe, Roskilly, Hudson, Golabek and McNutt2013). However, not everyone wants to become an expert in navigation technology. Similarly, those developing new navigation sensors and techniques would like to assess their performance as part of an integrated system without having to become experts on every subsystem. Thus, the research community needs modular navigation and positioning technology that enables the easy modification of individual hardware and software components. Hardware can be shared with consumer and professional user equipment. However, the software would ideally be open source to enable easy modification and minimise costs. This is not necessarily compatible with the business models of the commercial companies supplying consumer and professional users. Therefore, the research community itself would have to develop this software resource.

2.4. The Military

For military applications, solution availability is the key performance requirement. There must be a useable position solution regardless of context. In particular, very high host vehicle dynamics and hostile jamming of GNSS and other radio signals must be tolerated. Some applications require indoor, underground or underwater operation. Consequently, there is great interest in using environmental features and signals of opportunity to maximise robustness (Miller et al., Reference Miller, Uijt de Haag, Soloviev, Veth and Raquet2009). The Defense Advanced Research Projects Agency (DARPA) has proposed the term “all-source” to describe complex multi-sensor positioning and navigation systems (DARPA, 2010).

For most military applications, the accuracy required is a few metres, while the integrity requirement varies. Reliability is not critical for applications such as guided weapons and dismounted personnel, as an element of risk is assumed, but it is much more important for high-value platforms, such as ships and aircraft. Applications such as air-to-air refuelling and automatic landing of aircraft on ships require sub-metre positioning with a high level of integrity (Groves et al., Reference Groves, Offer, Mather, Pulford, Ashokaraj and Macaulay2008). Cost, size, mass and power consumption requirements also vary between applications; generally, the larger the host vehicle, the larger these may be.

The consumer model of a single device that is expected to perform both personal and vehicle navigation is unlikely to extend to the military. This is not only due to the differences in requirements. To maintain situational awareness, all military platforms, vehicles and dismounted personnel must have their own navigation systems that must communicate securely with each other. The communication links provide an ideal platform for cooperative (or collaborative) positioning that can enhance robustness in challenging environments through the sharing of information between peers (Grejner-Brzezinska et al., Reference Grejner-Brzezinska, Toth, Gupta, Li and Wang2010). This requires all equipment to be mutually compatible, which would be facilitated by a common modular integration architecture.

There are a number of key differences between military and civilian user requirements. The first is the military requirement for stealth. Vehicles and personnel must not emit signals that enable them to be detected by their opponents. Essential communications links use spread spectrum techniques and/or narrow beams to avoid detection. This places constraints on the use of active sensors such as radar and remote positioning techniques, whereby an object is positioned using the signals it emits. Secondly, military components must typically exhibit ruggedness in the form of greater tolerance to physical shock, temperature extremes and electromagnetic radiation, increasing their cost. A third military requirement is technological advantage over opposing forces. Consequently, many technologies developed under military funding are restricted in their use, so are not available for civilian applications until other countries have duplicated the technology. These restrictions have applied to many types of inertial sensor, including the new inertial sensor (and clock) technology that DARPA is investing in (Shkel, Reference Shkel2011). Finally, the military require the ability to jam signals used by their opponents without jamming their own signals, which is reflected in the current movement towards using separate GNSS spectrum for military and open-access signals. For all these reasons, military and civilian navigation hardware will largely continue to evolve separately. However, there is scope to share software modules and the overall modular framework.

3.1. GNSS

The manufacturing cost of GNSS chipsets has now dropped sufficiently to enable high performance receivers to be manufactured at low cost. This renders feasible a “universal” GNSS receiver that is suitable for all user groups and applications with the caveat that separate civilian and military versions may be needed. However, there remains a trade-off between performance and power consumption, which makes it a challenge to reconcile the requirements of consumers and professional users. One option is a receiver design that may be reconfigured to meet the needs of different applications by varying the number of front ends in use, the sampling rates of the analog-to-digital converters and the number of correlation channels in use, all of which affect the power consumption (Shivaramaiah and Dempster, Reference Shivaramaiah and Dempster2011). The number of correlators per channel and their spacing, the tracking bandwidths, acquisition algorithm and positioning algorithm would also need to be adjustable. For the lowest cost consumer applications, the manufacturing cost of the front ends is significant. One way of addressing this is to produce a single-front-end core receiver chip to which additional front-end chips may be connected (Mattos, Reference Mattos2013).

A common GNSS antenna for all applications is not practical, however. For consumer applications, the antenna must be small and cheap, particularly for smartphones. For most professional applications, the antenna sensitivity, polarization discrimination and phase centre stability take priority, which requires a larger, heavier and more expensive antenna.

3.2. Other Radio Positioning Technologies

Different non-GNSS radio signals are used for different applications (Groves, Reference Groves2013a). Some examples include distance-measuring equipment (DME) for aviation, enhanced long-range navigation (ELoran) for marine navigation, Wi-Fi for indoor and urban positioning, and Locata for land surveying. By combining a broadband antenna with one or more tuneable front ends with variable sampling rates, it is possible to construct a single set of hardware that can operate with many different types of positioning signals and signals of opportunity (Mathews et al., Reference Mathews, Macdoran and Gold2011). This device could also be used to receive GNSS. Because of its much lower frequency, ELoran requires a separate antenna and front-end. However subsequent signal processing can use shared hardware (Mattos, Reference Mattos2009).

3.3. Inertial and Magnetic Sensors

Inertial sensors that both exhibit high performance and are low-cost, small, light and low-power do not exist at present. A modular navigation system must therefore use different grades of sensor for different applications. However, for magnetic sensors, a common hardware design is feasible because the heading errors arising from magnetic model limitations, hard- and soft- iron equipment magnetism and environmental magnetic anomalies greatly exceed those arising from the sensors (Groves, Reference Groves2013a).

Considering the processing algorithms for inertial and magnetic sensors, inertial navigation, accelerometer levelling and magnetic heading algorithms can operate with almost any sensors and in most environments. However, pedestrian dead reckoning using step detection, host vehicle motion constraints, zero velocity updates and zero angular rate updates all depend fundamentally on the behavioural context (Groves, Reference Groves2013a). These context-dependent algorithms can substantially improve performance under the right conditions. Therefore, a truly universal approach to processing inertial sensor measurements must be context-adaptive, detecting the behaviour of the host vehicle or user and adapting accordingly (Groves et al., Reference Groves, Martin, Voutsis, Walter and Wang2013c).

3.4. Cameras

A basic digital camera of the type found in any smartphone or webcam is useful for almost any navigation and positioning application. For some applications, it would be useful to supplement this with an infrared camera. The hardware challenge lies in ensuring a camera is pointing in the appropriate direction, which is not always achieved by attaching it to the user interface, as is the case for smartphones. This leads to the question of whether the camera should be mounted separately, in which case the lever arm separating it from the other navigation sensors must be determined, or whether the rest of the navigation system should be attached to the camera, in which case the user interface is likely to be separate. In either case, power and data must be conveyed between the two parts of the system. In practice, multiple cameras, pointing in different directions and/or providing stereo imaging are both feasible and useful for many applications.

A camera can be used for both position fixing, by comparing image features with a database, and for dead reckoning, by comparing successive images to infer the user motion (Groves, Reference Groves2013a). Determining how to process the images depends on the context; for example, the camera may be downward looking for air applications and forward looking for land navigation. For image-feature matching, there will be different databases, containing different classes of feature, for different environments. Note also that an approximate position solution is needed to determine which region of the feature database to search and, if necessary, what to download from a server. Finally, image-based navigation remains an active research topic and is more advanced for some applications than others. Therefore, different image-processing software will be required for different applications.

3.5. Other Technologies

Most other navigation and positioning technologies are application-specific. For example, sonar and acoustic ranging systems are usually designed to work underwater, while wheel speed sensors are only applicable to wheeled vehicles. Radar and laser sensors may be used on a number of different vehicles. However, air, land and marine implementations are different and the cost, size, mass and power consumption of these sensors is too high for many navigation and positioning applications. The sensors must also be mounted in the right place on their host vehicles.

In a modular navigation system, application specific sensors could be incorporated on a plug and play basis if a suitable interface were to be provided. This could be based on the universal serial bus (USB) or Bluetooth. Depending on the overall system architecture, the information supplied could be sensor measurements, position and velocity information or ranges and range rates.

Map matching can be used for most land navigation applications and does not require additional hardware, other than for storing data. However, the implementations for road vehicles, trains and pedestrians are quite different (Groves, Reference Groves2013a). Therefore, the context must be known for it to be implemented correctly.

4.1. Hardware Architecture

There are a number of different ways of constructing a modular navigation system depending on the trade-off between development and production costs. One approach is to adopt a standard core that is augmented with different peripherals for different applications. Figure 1 illustrates this. The hardware core could comprise a reconfigurable GNSS receiver, a reconfigurable general radio, low-cost inertial and magnetic sensors, a processor and memory. A variation on this is to use the processing capacity and memory of a host device. There could be a built-in camera and/or connectors for external cameras could be included. Different antennas would be connected for different applications and a series of interfaces provided for connecting application-specific sensors, as required. Higher quality inertial sensors could be connected via this interface or incorporated in higher performing versions of the hardware core.

Figure 1. Core and peripheral hardware architecture.

An alternative model would be to design a set of interfaces that enables a range of different hardware (and software) modules to be connected together to meet the requirements of any navigation or positioning application. This would reduce the production costs for low-performance applications such as consumer devices. However, with a wider range of hardware modules, the overall development cost would be higher.

4.2. Integration Algorithms

Any multisensor navigation or positioning system needs integration algorithms to obtain the best overall position solution from the constituent subsystems (Groves, Reference Groves2013a). However, there are a number of challenges that must be addressed. To obtain the best performance, integration algorithms must not only input measurements from a wide range of subsystems, but also calibrate systematic errors in those subsystems. However, designing the integration algorithms requires expertise in all of the subsystems, which can be difficult to establish in a single organisation. The move towards navigation systems with larger numbers of subsystems and modular designs that use different subsystems for different applications will make this problem worse.

An additional complication is that different modules in an integrated navigation system are often supplied by different organisations. These modules must work together. However, organisations can be reluctant to share the necessary design information as this is considered to be IP that must be protected. An extreme example of this is the smartphone: one company supplies the GNSS chip, another supplies the Wi-Fi positioning service, a third organisation supplies the mapping, the network operator provides the phone-signal positioning, a fifth company provides the inertial and magnetic sensors and a sixth company produces the operating system. Due to lack of cooperation between these different organisations, useful information gets lost. For example, GNSS pseudo-range measurements are not normally available to “app” developers and there can be gaps in inertial sensor data.

With new innovations constantly being introduced, a method is also needed for incorporating new positioning technologies and methods within integrated navigation systems without having to redesign the integration algorithms each time something is added.

Finally, and most fundamentally, an integrated navigation system is not truly modular if a bespoke integration algorithm is required for each application. For all of these reasons, a method of modularising the integration algorithm is required. One solution is to divide the integration algorithm into a universal integration filter module and a series of configuration modules, one for each subsystem. Figure 2 shows an example.

Figure 2. Example modular integration of GNSS, inertial navigation and visual navigation for an air application (each colour denotes a separate supplier).

The integration filter module would be designed by data fusion experts and would accept a number of generic measurement types, such as position fixes, pseudo-ranges and inertial sensor measurements with associated metadata. Unlike conventional integration filters, a universal filter must reconfigure its measurement vector, state vector and associated matrices according to the measurements available. This capability is sometimes called “plug and play” and a number of prototypes have already been developed (Hide et al., Reference Hide, Moore and Hill2007; Penn, Reference Penn2012; Soloviev and Yang, Reference Soloviev and Yang2013). Table 1 lists the possible measurement types and their associated positioning methods. A truly universal integration filter should be able to support measurements in the position and velocity domain (i.e., loosely-coupled integration) and in the range and angle domain (i.e., tightly-coupled integration). It should also be able to support both total-state integration, whereby the navigation solution is directly estimated by the integration algorithm and error-state integration, whereby the errors of a reference navigation solution are estimated. The fundamental positioning methods and the different types of integration architecture are explained in Groves (Reference Groves2013a).

Table 1. Fundamental measurement types and their associated positioning methods.

¹ Resolved with respect to an external coordinate system, such as north, east, down.

² Requires an antenna array.

³ Or the integral thereof.

⁴ Differencing across receivers may be performed by the relevant configuration module prior to input by the universal filter module.

⁵ Differenced across transmitters or environmental features.

The integration filter could be an extended Kalman filter, an unscented Kalman filter, a particle filter, or a new type of data fusion algorithm. Different filter designs are suited to different applications. There are different ways of handling large uncertainties, unreliable measurements and irregular error distributions. Kalman filter-based estimation algorithms have maximum error tolerances, beyond which significant modelling errors can arise. For particle filters, there is a trade-off between the number of subsystem errors estimated and the complexity of the error distributions that can be handled. Overall, there is a three-way trade-off between robustness, processor load and precision.

The configuration modules would convert the subsystem outputs into the standard measurement types and provide the metadata necessary to integrate those measurements. This comprises the locations of transmitters and other landmarks, which subsystem errors should be estimated by the integration filter module and statistical descriptions of the measurement errors and subsystem errors, including their variances and the power spectral density of their time variation. This statistical information is essential to ensure correct weighting of the measurements and enable position uncertainty bounds to be computed. Thus, for example, to incorporate inertial sensor measurements, the inertial sensor specifications are required. For each measurement type, a standard set of subsystem errors that could be estimated would need to be agreed, such as biases, scale factor errors, temperature dependencies and range ambiguities. The correlation across multiple measurement streams of ranging biases due to clock offsets must also be specified. For each error type, both open-loop and closed-loop correction (Groves, Reference Groves2013a) should be supported.

Each configuration module would normally be provided by the supplier of the subsystem to which it applies. The developer of a new navigation or positioning technique could incorporate their subsystem within an existing integrated navigation system simply by writing a configuration module for it; they would not need to become experts in sensor integration or the other subsystems. It is tempting for developers to produce configuration modules that provide over-simplistic and/or over-optimistic descriptions of their subsystems. Therefore, some form of certification procedure would be needed to enable other organisations to trust a new module.

For total-state integration, a dynamics model configuration module would also be needed to provide the universal filter with information on the host vehicle (or pedestrian) dynamic motion.

The key to the success of this modular integration approach is the definition of the interfaces between the universal filter module and the configuration modules. With a clearly defined and open interface standard, it should be possible to combine any integration filter module with any set of configuration modules without the module designers needing to know the inner workings of the other modules. This leaves organisations free to protect the IP within their own modules. However, it is still necessary to ensure that the filter module is suitable for the application in question. Furthermore, in practice, a filter module may not be designed to support every measurement type and there will be limits to the sizes of the errors it can handle. The interface standard must therefore incorporate a capability specification for each filter module and a protocol should be developed for handling mismatches between the measurements and filter module. A certification process would also be needed to ensure that the filter module actually has the capabilities it claims. However, it is primarily the responsibility of the overall system designer to ensure that a suitable filter is selected.

4.3. System Reconfigurability

In a modular navigation system, hardware and software modules would be expected to operate under different environmental and behavioural contexts for different applications. The optimum design of the signal processing algorithms for GNSS and other radio positioning systems depends on the dynamics and reception conditions. Different image-based navigation and positioning algorithms are also suited to different environments and system dynamics, while the integration algorithm's dynamics model will always depend on context. A modular navigation and positioning system must also be able to adapt to the differing user requirements of different applications, such as accuracy, integrity, solution availability, update rate and power consumption.

Therefore, a modular navigation system must be reconfigurable. This can be achieved by adding a system control module that specifies the context and user requirements. This is then passed via the open-standard interface to the configuration modules, which configure both the integration filter module and the subsystems themselves accordingly. Some subsystems will require different processing algorithms to be selected for different contexts and/or requirement sets. It may even be necessary to activate or deactivate subsystems according to the context.

The requirements should also be input directly to the filter module. In practice, different filter modules are likely to be needed, depending on whether high accuracy, high integrity or high solution availability take priority. This is because different design philosophies have historically been applied to meet these different sets of user requirements and significant effort would be required to reconcile them.

For applications, such as consumer navigation, where the context changes as the system is used, this can be detected using the available sensors and the system reconfigured accordingly. This is known as context-adaptive navigation (Groves et al., Reference Groves, Martin, Voutsis, Walter and Wang2013c). Many navigation and positioning systems must also be able to switch between the requirements of different applications. For example, a consumer device may be used for navigation that requires continuous positioning, for tracking which requires position updates at regular intervals, and for location-based enquiries that only require single position fixes. Similarly, a single professional positioning device might be used for surveying static points, for setting out which requires the system to be moved until the desired location is found, and for controlling construction machinery which requires continuous positioning.

Figure 3 shows a suitable system architecture for a context-adaptive multi-application reconfigurable navigation or positioning system. Note that further research on context adaptivity and integrity frameworks for complex multisensor systems is needed before system reconfigurablility can be realised in full.

Figure 3. System architecture for a context-adaptive multi-application reconfigurable navigation system (each colour potentially denotes a separate supplier).