NAVISP'S WEBINAR SPARKS INTEREST IN THE CAPABILITIES OF IMAGING SENSORS AND 3D CITY MODELS TO ASSIST GNSS IN MOBILITY APPLICATIONS

The NAVISP Webinar "Assisted GNSS with Imaging Sensors and 3D Models for Mobility Applications", whose objective was to launch the related NAVISP Thematic Window, was held on the 6 th of July 2021 with a large participation from navigation experts, industry, academia and other representatives from economic operators (please follow the links to read the presentation slides and see the video recording of the Webinar).

As an introduction, ESA representatives explained the objective of the related NAVISP Thematic Window detailing its benefits and the mechanism of how to apply to it (for more information on the call and how to submit a proposal please click here for Element 2 and here for Element 3). ESA is looking forward to receiving proposals submitted to NAVISP by the 31 st of October 2021.

Urban mobility requires ubiquitous, reliable and resilient PNT solutions, but GNSS in general cannot cope with all the challenges of the urban environment. Augmentation technologies can help overcome these challenges.

To explain how, the Webinar included presentations from a panel of external experts.

Floris van de Klashorst, from Spacetec Partners, showed that, from a market perspective, the introduction of advanced technology in urban mobility is a very promising area due to the benefits in improved efficiency and reduced emissions.

He moved from the global mobility trends of megacities and CO2 emissions and energy per capita reduction or control, to predict that no single transport mode will answer all the questions. Instead, new transport modes versus existing, public versus private, shared versus dedicated, can all offer unique benefits for different cases.

Waterway's mobility offers unique value compared to road mobility. Waterway’s mobility benefits from other innovations, accelerating related implementations. Self- steering vessels autonomy is more easily implemented and vessels can reuse partially depleted vehicle batteries, due to the lower speed required.

Air Mobility leverages the sky to create networks between people, cities and regions. The key use cases are Delivery (express), Passenger, Cargo, Mission Specific (Medical, Surveillance, Defense). Air mobility complements multimodal mobility system regionally, but remaining concerns are safety in the air (and on the ground) and noise pollution.

As a conclusion, Urban Mobility trends show an integrated multimodal system of many modalities with people-oriented cities. Cities have to orchestrate integrated mobility systems to balance demand and supply and create a competitive market.

Paul Groves, from University College London, explained the challenges for GNSS of dense urban areas. Conventional GNSS positioning algorithms assume a direct path from each satellite to the receiver antenna, whereas signals are in reality blocked, reflected and diffracted by buildings and vehicles. He showed that 3D mapping and imaging sensors are becoming some of the most suitable ways of augmenting GNSS in urban environment.

With 3D-Mapping-Aided (3DMA) GNSS, projection techniques can rapidly predict which satellites are directly visible at multiple locations. Ray-tracing techniques can predict the path delay of Non-Line-of-Sight (NLOS) signals, enabling correction of NLOS errors, and predict which direct Line-Of-Sight (LOS) signals may be subject to severe multipath.

Since 3DMA methods need an initial position to predict satellite visibility, sophisticated algorithms are needed. For example, likelihood-based 3DMA ranging sets up a grid of candidate positions, scores each position according to how consistent the measured and predicted pseudo-ranges are and derives the position from these scores. Consistency is scored in a basic way by using an asymmetric distribution for signals predicted to be NLOS or in a more advance way by using ray tracing to correct NLOS signals.

Another technique is Shadow Matching, whereby a grid of candidate positions is scored by comparing the measured signal-to-noise (instead of pseudo-ranges) and predicted satellite visibility. While ranging is better for along-street positioning, shadow matching is better for across-street positioning, and both can be combined in an integrated technique.

Regarding imaging sensors, many different types can be used for navigation and positioning (e.g. visual and infrared cameras, laser scanner, LIDAR).
With image-aided GNSS signal selection, visual and infrared cameras pointing upwards provide wide-angle sky images showing buildings that can be used to determine which satellite are visible and which are not, thereby excluding NLOS signals.

Alternatively, with visual odometry, the cameras can point forward and provide a series of images, from which features are extracted to infer the user velocity. Velocity can be integrated and merged with position from GNSS, in a similar way to inertial navigation.
Further, with image-based absolute positioning, the position is determined by comparing features in a live image with those in a database (with GNSS used to determine the search area).

Sonwya Gopal, from the map-making company HERE, presented the concept of maps for humans and machine, as high fidelity 3D maps with rich features to support AGNSS are becoming widely available.
Users desire visual cues that help orient themselves to their surroundings (e.g. to answer the question: at what building do I make the turn?).
Cars on the road today, like the drivers, need to know where they are, which way they go, what is around them, how they should behave and how reliable the data is. Modern maps answer most of human and machine's questions. For example, for
machines, maps include a description of the roads which features lane topology,
geometry and other useful attributes that can be used in automated processing.

Concluding the webinar expert presentations, Alison Lowndes, from the company NVIDIA, made the case that increasing processing power at user level and in the cloud, and applying Artificial Intelligence technology, are making advanced and computationally intensive solutions possible.
She introduced a new paradigm for Data Centres, based not only on Central Processing Units (CPU) and Graphical Processing Unit (GPU), but also on Data Processing Unit (DPU), specialised System-on-Chip able to handle data at the level of the network directly, enabling for example applications linked to high data rates like in 5G. She also presented compact and powerful processing platforms for Users.
This processing power, distributed conveniently and according to the application between Data Centre and User, can be harnessed by specifically designed processing libraries to perform trajectory clustering based from images, or real time image recognition from traffic camera on fixed or movable platforms.
After having shown the practical application of tensors calculus (the same context where general relativity is developed), in deep learning implemented in GPUs for industrial motion control applications, she concluded highlighting the benefits of using a simulation environment to accelerate the training of autonomous cars Artificial Intelligence algorithms.

The webinar ended with a round of Questions & Answers, moderated by ESA representatives, which showed the high interest in the community around Assisted GNSS with Imaging Sensors and 3D Models for Mobility applications.