Artificial Intelligence in GNSS Receivers

Register for the Final Presentation of the NAVISP EL1 034 Project: "AI-enabled baseband algorithms for high-fidelity measurements (AIGNSS)"

In the context of the increasingly stringent requirements on GNSS performances for various business applications such as autonomous vehicles (in particular cars), several technological gaps with current positioning techniques need to be addressed. Current technologies based on numerical models and empirical parameters are not efficient in harsh environment (Urban canyon, low C/N0…). In these specific areas, one solution consists in taking into account a priori knowledge of the environment (such as 3D maps) correlated to the GNSS signal received. 

This raises several challenges: 

• A high refresh rate of this a priori knowledge to avoid any fluctuations (new building, tall vehicle blocking the LOS) which make a priori knowledge obsolete,
• High computational and memory resources need,
• Highly adaptive configuration and parameters to cope with the variability of the autonomous vehicle environment.

This is where AI is expected to play a role in filling these Gaps. 

The AIGNSS exploratory project investigated a new paradigm in the design of GNSS algorithms, leveraging artificial intelligence to design and trial algorithms able to enhance raw measurements, along with quality indicators related to the local environment.

The AI algorithms were tested on different GNSS datasets, covering several kinds of data and hybridization scenarios. Most promising algorithms were selected not only in terms of global added PVT accuracy but also regarding their overall efficiency and performances in specific use cases.

The AIGNSS key activities are summarized as: 

• A review of the state of the art on the AI enhanced GNSS receivers. 
• A flexible and modular AI GNSS testbed allowing to investigate, test and benchmark various AI algorithms, both ones implemented in the frame of this activity and new ones. This testbed shall also allow for the plugging of any AI algorithm to the desired data input (input signal, correlator output, raw data, PVT…) 
• A collection of relevant training data (both simulated and real) 
• An implementation of three selected AI algorithms on the testbed 
• An analysis & benchmark of these three algorithms based on extensive field trials.

The project was carried out in the scope of NAVISP Element 1, which is dedicated to technology innovation of the European industry in the wider PNT sector.
 

Please check the Agenda and register for the event!