In order to answer the main challenges of sustainable development, transport systems need to change towards greener solutions, as well as being reliable and smart. This leads to the development of autonomous vehicles among other solutions. These new systems rely on different smart sensors, among which Global Navigation Satellite Systems (GNSS) receivers are considered a core element. GNSS can potentially offer global continuous positioning, but suffers from errors and may contain the presence of multiple threats at both system and local user level. Moreover, degraded performances are expected in urban areas due to the presence of obstacles in the vicinity of the vehicle that reflect or diffract the signals and impact the time of arrival measurement. This compromises therefore the global positioning performance requirements such as accuracy, availability and integrity. A good knowledge of these effects must help increase these performances at GNSS level but also for the development of fail-safe multi-sensor solutions as requested by most of the mobile users (road, railway, UAVs).
These effects can be modelled by statistical or deterministic tools but the fact is that real behaviour in a real urban environment is complex to model and depends on a large number of parameters that make it difficult to recreate and completely model.
With the development of artificial intelligence for data modelling and classification come new opportunities for GNSS service provision and localization evaluation or enhancement. The goal of this collection is to merge the different contributions using AI in GNSS-based localisation that will improve knowledge, characterization and performance.
First uses show the interest of machine learning for satellite visibility or satellite state of reception definition (LOS/NLOS), multipath detection and interference detection. New concepts based on context-detection are also investigated that will allow the development of adaptive solutions. Context can be analysed as signal reception environment but also as user behaviour. Such a knowledge will be used for precision enhancement, but also for sensor fault detection and robustness improvement. In addition, the signal spoofing and interference is another threat that can degrade the performance of GNSS based localization. Based on the ambiguous nature of signal interference and attack, it is also very challenging to detect.
Furthermore, new system and signal monitoring techniques can be developed thanks to the potentially large available data from ground networks, which is an important requirement in AI-based anomaly detection methods.
AI applications to such a system remains quite recent and is increasing rapidly. This collection expects to be a place for sharing exploration of these new solutions.
The following topics have been identified but proposals are not limited to this list:
- Building datasets: A certain number of solutions require supervision. However, availability of some labelled dataset is an important challenge. The use of simulated datasets could be one of the solutions. With experimental data, one of the issues is to determine references to label real datasets in complex environments, in particular when dealing with dynamic trajectories of vehicles.
- New uses of AI for better GNSS-based localisation performances
- Refined context detection
- Error models and reception characterization, in particular multipath characterization and detection in challenging GNSS environments.
- Spoofing and interference detection based on artificial intelligence
- AI-based anomaly detection for signal monitoring (e.g., signal deformation)
In order to answer the main challenges of sustainable development, transport systems need to change towards greener solutions, as well as being reliable and smart. This leads to the development of autonomous vehicles among other solutions. These new systems rely on different smart sensors, among which Global Navigation Satellite Systems (GNSS) receivers are considered a core element. GNSS can potentially offer global continuous positioning, but suffers from errors and may contain the presence of multiple threats at both system and local user level. Moreover, degraded performances are expected in urban areas due to the presence of obstacles in the vicinity of the vehicle that reflect or diffract the signals and impact the time of arrival measurement. This compromises therefore the global positioning performance requirements such as accuracy, availability and integrity. A good knowledge of these effects must help increase these performances at GNSS level but also for the development of fail-safe multi-sensor solutions as requested by most of the mobile users (road, railway, UAVs).
These effects can be modelled by statistical or deterministic tools but the fact is that real behaviour in a real urban environment is complex to model and depends on a large number of parameters that make it difficult to recreate and completely model.
With the development of artificial intelligence for data modelling and classification come new opportunities for GNSS service provision and localization evaluation or enhancement. The goal of this collection is to merge the different contributions using AI in GNSS-based localisation that will improve knowledge, characterization and performance.
First uses show the interest of machine learning for satellite visibility or satellite state of reception definition (LOS/NLOS), multipath detection and interference detection. New concepts based on context-detection are also investigated that will allow the development of adaptive solutions. Context can be analysed as signal reception environment but also as user behaviour. Such a knowledge will be used for precision enhancement, but also for sensor fault detection and robustness improvement. In addition, the signal spoofing and interference is another threat that can degrade the performance of GNSS based localization. Based on the ambiguous nature of signal interference and attack, it is also very challenging to detect.
Furthermore, new system and signal monitoring techniques can be developed thanks to the potentially large available data from ground networks, which is an important requirement in AI-based anomaly detection methods.
AI applications to such a system remains quite recent and is increasing rapidly. This collection expects to be a place for sharing exploration of these new solutions.
The following topics have been identified but proposals are not limited to this list:
- Building datasets: A certain number of solutions require supervision. However, availability of some labelled dataset is an important challenge. The use of simulated datasets could be one of the solutions. With experimental data, one of the issues is to determine references to label real datasets in complex environments, in particular when dealing with dynamic trajectories of vehicles.
- New uses of AI for better GNSS-based localisation performances
- Refined context detection
- Error models and reception characterization, in particular multipath characterization and detection in challenging GNSS environments.
- Spoofing and interference detection based on artificial intelligence
- AI-based anomaly detection for signal monitoring (e.g., signal deformation)
