Tag: AI

UAV and AI update
A couple of stories about unmanned air vehicles in the war in Ukraine and a response to the recent Open Letter by the “Future of Life Institute” with more than 200,000 signatures on advanced AI, which urged a six-month moratorium to allow the development of seemingly much needed AI regulations.

The war in Ukraine

It has been reported that Ukrainian forces were operating the commercially available Chinese Mugin 5 UAV, presumably for surveillance of Russian forces inside Russian-occupied territory. The Mugin 5 can be bought commercially for $10-15,000 and is manufactured by Mugin, which is based in the port city of Xiamen, on China’s eastern coast. In a previous statement posted on the company’s website on March 2, Mugin Limited said that it “condemns” the use of its products during warfare and that it ceased selling products to Russia or Ukraine at the start of the war. However, Russian forces claimed in January 2023 that it had actually shot down one of these Chinese-made UAVs being flown by Ukrainian forces over their territory.

Then, just this week, Ukrainian forces apparently were able to track a low level, slow-moving air vehicle coming at them from Russian occupied territory. After some time, they were able to intercept the UAV, which carried a flashing navigation light, from the ground, and were able to bring it down using small arms. The remains of the crashed UAV were found in a clearing in the forest; a single 44 lb bomb was removed from the wreckage and safely exploded by the Ukrainian team.

Weaponized Mugin 5 following crash in Ukraine forest. (Image: Screenshot from video from Kanal13 Youtube)

Somewhat worse for wear, the Mugin 5 UAV appears to have been held together in places by duct tape and other patches. Is it possible that having shot down a Ukrainian surveillance UAV the Russians recovered these remains and crudely restored the unit to flying and navigating capability, then sent it back to Ukraine owners carrying a bomb? Anything is possible in this conflict.

Staying with this conflict and the use of UAVs by both sides, its seems that Australia has come up with a low-cost surveillance UAV that is virtually undetectable and it’s proving quite popular with the Ukrainians. Most defensive detection involves some form of radar scanning, which relies on radar returns bouncing off a flying target. The Australian company SYPAC in Melbourne has developed the Corvo Precision Payload Delivery System (PPDS). It is a wax-coated cardboard UAV, held together with elastic bands and glue, but carrying sophisticated guidance and control electronics.

(Image: Screenshot of video posted by 7 News Australia)

SYPAQ has developed the CORVO UAV under an AU $1.1 m government contract with the objective of creating a low-cost, disposable UAV to deliver urgent needs — such as medical supplies or to resupply small arms ammunition to the Australian military. CORVO is autonomous once launched, using GNSS guidance, or dead reckoning if GNSS signal is lost or jammed. Apparently, hundreds of these disposable UAVs have already been shipped to Ukraine.

While a surveillance role was originally envisaged in Ukraine, it is reported that, “They have been very good at inflicting lots of damage on the enemy,” according to Ukraine’s ambassador to Australia. So, CORVO UAVs may well have already been weaponized.

Open Letter on AI development

Following a recent open letter supported by Elon Musk and Steve Wozniak that proposes a six-month halt on advanced AI development, I was recently approached on behalf of Professor Ioannis Pitas, director of the Artificial Intelligence and Information Analysis (AIIA) lab at the Aristotle University of Thessaloniki (AUTH) and management board chair of the AI Doctoral Academy (AIDA) with somewhat different views.

In order to further the on-going discussion, I thought it would be appropriate to give some space to an alternate view on AI development. So here are some paraphrased comments approved by Pitas:

Could AI research be stopped even for a short time? It is doubtful. Further AI progress is necessary for us to transition from an information society to a knowledge society.

Maybe we have reached the limits of AI research carried out primarily by Big Tech, which appears to treat powerful AI systems as black boxes whose functionality may be poorly understood.

It seems that the open letter reflects welcome and genuine concerns on social and financial risk management. Are expensive lawsuits in an unregulated and unlegislated environment inevitable as a consequence of ill-advised AI pronouncements?

However, it is doubtful whether the proposal for a six-month ban on large-scale experiments is the solution. It’s impractical for competitive commercial and geopolitical reasons, with very few benefits.

Of course, AI research can and should become more open, democratic and scientific.

Here are a number of suggested options:
- Should elected parliaments and governments make the important decisions on AI rather than corporations or individual scientists?
- Every effort should be made to facilitate the positive aspects of AI social and financial progress and to minimize any negative aspects.
- The positive impact of AI systems can greatly outweigh their negative aspects if proper regulatory measures are taken.
- It is possible that the biggest threat is that AI systems could deceive too many people who have little related knowledge. This can be extremely dangerous.
- We should counter the big threat coming from the use of AI in illegal activities — cheating on university exams is a rather benign use — while the possibility of criminal exploitation may be very much worse.
- The impact of AI on labor and markets will be very positive in the medium to long term.
- AI systems should be required by international law to be a) registered in an ‘AI global register’, and b) users should be notified when they converse with or use the results of an AI system.
- As AI systems have a huge impact on society, and in order to maximize their benefit and socio-economic progress, it is recommended that:
  o advanced key AI system technologies should become mostly open
  o AI-related data should be at least partially open.
- However, strong financial compensation schemes should be established now for AI technology developers to compensate them for any component that becomes open source.
Well, this is a bit of a departure from our nominal UAV/AI report, but there does seem to be a growing number of voices calling for some form of AI regulation and more extensive discussion might well help this movement come to a conclusion. And it would seem that the U.S. administration is listening, as the U.S. Commerce Department has announced that it is seeking inputs from interested parties for methods to test the safety of AI systems — to ensure that they are “legal, effective, ethical, safe and otherwise trustworthy.” In order to enforce these standards, the department is investigating whether audits and inspections to certify AI systems should be required before their release on the unsuspecting public.

The U.S. Commerce Department is apparently not alone in these concerns, as China is also looking to ensure that systems such as Alibaba Cloud’s Tongyi Qianwen, a competitor to OpenAI’s ChatGPT, are socially beneficial. Meanwhile, following the release of ChatGPT and similar products from Microsoft and Google, awareness has grown of the capabilities of the latest AI tools that generate human-like text passages, and even new images and video. The UK Department for Science, Innovation and Technology and the Office for Artificial Intelligence on the other hand, seem to be looking for an approach to regulation that will not restrict AI innovation.
April 19, 2023
Geospatial and location intelligence capabilities highlight GCA 2023

Geo Connect Asia (GCA) 2023, Asia’s leading international geospatial industry event, will take place March 15-16 at Marina Bay Sands Expo and Convention Centre, Singapore.

GCA 2023 will be held alongside Digital Construction Asia (DCA) 2023, and co-located with the launch of Drones Asia 2023. The three-in-one event, held fully in person, is expected to bring together more than 2,500 delegates and attendees from around the world.

With the theme “Advancing sustainable and resilient geospatial solutions for an interconnected world”, a key focus of GCA 2023 will be the use of advancements in geospatial technology and data interoperability to address regional challenges.

Supported by the Singapore Land Authority (SLA), the event will feature more than 70 exhibiting companies and demonstrate the role played by the mix of geospatial, location intelligence, remote sensing and drone-based solutions.

The two-day in-person conference comprises ten main sessions featuring more than 50 prominent industry speakers, panelists and moderators.

Shining light on opportunities for enhancing productivity in the construction world, DCA 2023 will focus on showcasing digitalized processes and improved workflows. By enhancing ground-based equipment with aerial capabilities and implementing technology — such as artificial intelligence (AI), building information modeling and internet of things — current challenges in construction can now be targeted via novel and more efficient approaches.
Drones Asia 2023 will address the commercial UAV industry. The newly launched and co-located show aims to create a focused platform for the complete drone ecosystem.

Drones Asia 2023 plays a critical role in enabling AI in today’s geospatial marketplace, broadening the conversation as industry experts investigate the application of UAVs in the commercial and industrial world, exploring industrial adoption to improve productivity and efficiency.

For the full programme and registration, visit the GCA 2023 website.

March 14, 2023
Unmanned and Autonomous Systems for Utilities and Energy conference to be held virtually and in person

The 6th Unmanned and Autonomous Systems for Utilities & Energy Conference will take place in Atlanta, GA on June 8-9, 2022.

The event aims to provide a platform for UAS professionals to gain insight from industry peers and regulatory bodies on best practices in pilot training, safety in inspections, data management and security, updates on Part 107, new UAS technologies, and other key issues for utilities.

The conference also includes discussions on alternatives to foreign-made drones, BVLOS waivers and use cases, a closer look at LiDAR and AI, and building and refining drone programs to boost efficiency and reliability. Attendees will learn how they can navigate through industry challenges by leveraging emerging technologies and improving existing strategies to boost operational success.

Join the event to learn how you can navigate through industry challenges by leveraging emerging technologies and improving existing strategies to boost operational success.

Those who are unable to attend in person have the option to attend virtually. The Live+ content platform will give you access to all the presentations and is loaded with features to ensure full participation.

AUVSI members are entitled to a discount on full price conference fees (not valid for solution providers).

Learn more about the event at https://www.marcusevans.com/conferences/unmannedsystems.

For registration information, cost and any discounts that may apply please contact:
Ria Kiayia
Digital Media & PR Marketing Executive
[email protected]

May 24, 2022
Registration open for Munich Satellite Navigation Summit
The Munich Satellite Navigation Summit program is now online and registration for the event is open. The summit will be held online March 7-8.

The Munich Satellite Navigation Summit focuses on satellite navigation in the present day and future, featuring global speakers and highlighting the latest developments in the field of GNSS. This year’s theme is “AI in GNSS – Intelligence brought to Navigation”.

The summit will feature 12 sessions from industry experts, including sessions on the following topics:
- First and Second Generation of the European Satellite Navigation System Galileo
- Modernization of the US Global Positioning System
- Status and modernization of the Russian Global Satellite Navigation System GLONASS and the Chinese Beidou System (BDS)
- Developments of regional systems like the Japanese QZSS and the Indian IRNSS and the Korean Positioning System (KPS)
- Use of AI within the navigation world and its implications
- Jamming, spoofing, interference, and countermeasures; understanding secure Galileo services (OSNMA, PRS)
- GNSS and the new race to the Moon; upcoming space mission related to PNT
- Advanced technologies for PNT (quantum, optical) even beyond Galileo 2nd Generation
The summit will also offer a free job market discussion and company pitches prior to the main conference for all attendees.

To view the Munich Satellite Navigation Summit program and register, visit munich-satellite-navigation-summit.org
February 14, 2022
Unmanned and AI: Indy Challenge takes autonomous to big track

When I saw that there was a plan for a whole bunch of unmanned, semi-autonomous racecars to compete at the Indianapolis Motor Speedway (Indy, or IMS) racetrack, I initially thought we might be headed to one significant mess of broken-up machines and potentially a lot of damage. I tracked the various announcements of the competition as things progressed, especially when a prize of $1 million dollars was put up by the Lilly Endowment in Indianapolis, and the majority of the field appeared to be potentially staffed by undergrad university teams.

Photo: Indy Autonomous Challenge

However, this isn’t the first time we’ve had unmanned, autonomous road vehicles in competition — we’ve seen highly instrumented SUVs in desert settings in Nevada and California, initially with pretty poor results, which began to improve significantly for the second time round, then vehicles in some simulated street settings with some mixed and also some pretty good results.

So, as the competition date grew closer for the Indy Autonomous Challenge (IAC), the number of published progress reports began to increase, and we began to better understand how the initial 40 teams might take on this seemingly impossible task — how on Earth will they replicate a regular Indy (also a class of racecar) race? Surely many unmanned racecars on the same track at the same time doing more than 150 mph would be catastrophic!

When you take a look, however, at the advances we’ve seen, which have enabled unmanned cars, trucks, taxis and such – surely this tech could stretch to meet these major objectives? But Dallara AV-21 Indy Light racecars avoiding hurtling walls passing by, cornering, getting in and out of the pits, coping with vehicles behind, ahead and overtaking — even a superior-equipped unmanned racecar at >150 mph — well that’s something we would really need to see.

Then you have to take a look at the outfits involved, providing support to the IAC teams – companies including Cisco, and motor sport units such as ADLINK, Ansys, Aptiv, Bridgestone, Luminar, Microsoft and Valvoline and the non-profit Energy Systems Network. The University teams from around the world themselves appeared to also have significant heritage and skill-levels.

As the 40 University teams started the long trek to get over the hurdles that this challenge presented, members from 21 of those institutions were actually able to make it to Indy, grouped into nine “national” teams. By October 23 the nine teams, with only one car each, were ready to test their autonomous vehicles on the actual track.

Clemson University established the baseline Dallara AV-21 vehicle and technology to be used by each team for the race, with sensors monitoring chassis motion, suspension, tires and powertrain. Each team would install its own guidance and avoidance system, with each vehicle equipped with six cameras, four lidars, RTK GNSS, associated radios and bags of computing running each team’s customized control system software. The object being for cars to exit pit-lane, accelerate, brake, establish an optimum line for each corner and flat, avoid obstacles, evaluate the track conditions and establish tolerable limits.

The teams were required to complete several stages of selection, from submission of initial proposals through demonstration of existing vehicle automation capability, simulated race performance, qualification testing at the Indy track — all leading to an anticipated head-to head race against the other qualifiers.

Then 20 days of planned testing stretched to 50, and three months of preparation passed with students working intensely throughout, curing the glitches, experimenting with how to increase lap speed, and pushing the limits while still keeping the cars intact.

Energy Systems Network managed the rules of the final competition in a way that reflected Indy qualification days prior the main race — they judged that the technology was not yet at a stage where multiple cars on the track at the same time would have been such a good idea. So, each car was to individually run a number of practice/qualification laps and the quickest car would be the winner.

During the first stage of live competition, cars were required to exit the pits and run a warmup lap, followed by two laps that were timed and a slow-down lap that required navigating around inflatable barriers on the front-stretch, and then return successfully back around the track into their pit-stop locations. There were several spins in the corners and several crashes, but the four surviving cars/teams were able to optimistically post speeds of more than 130 mph.

Photo: Indy Autonomous Challenge

The final phase involved the four teams taking their cars around a number of warm-up/practice laps, followed by four timed laps. Only the car from Germany’s Technical University of Munich was able to complete all laps with an average speed of ~136 mph, so that team ultimately won the $1 million prize. Even so, all teams were able to successfully mature their systems’ performance through the many months leading up to the IAC and their progress through the various qualification stages. Even the other three final qualifiers had much to celebrate as a result of the competition.

The sponsors supporting the various teams as they progressed through the Challenge may have spent more than $120 million, so that high-pressure development work will be invested back into many vehicle automation opportunities. After all, that was the main objective for the whole undertaking. We should hopefully begin to see safer, more capable self-driving vehicles emerge in the months to come as the technology is applied to more production vehicle automation.

Tony Murfin
GNSS Aerospace

November 17, 2021
ESA: Baltic ferry gathers data for self-aware sailing

News from European Space Agency (ESA)

A day of ferry trips between Finland and Estonia became some of the best documented voyages in maritime history. Cameras, sensors, radio and satellite navigation receivers and even microphones recorded every instant of the crossings over the Baltic, gathering raw data for a new ESA-led project applying artificial intelligence (AI) to the situational awareness of shipping — as an important step to full autonomy.

The Tallink shipping company’s new 212.2 meter-long Megastar passenger and car ferry was fitted with data-gathering devices for its sailings on the busy stretch of sea between Helsinki and Tallinn.

The testing was overseen by a team from the Finnish Geospatial Research Institute (FGI) for an ESA project called Artificial Intelligence/Machine Learning Sensor Fusion for Autonomous Vessel Navigation, or Maritime AI-NAV.

“Our aim is to show how AI can be applied to achieve autonomous situational awareness, so that a ship can reliably sense its own environment,” said FGI’s Sarang Thombre.

Photo: European Space Agency

“Such autonomous systems would initially be deployed in support of human crews, for enhanced safety and efficiency – with crewless ships a much longer-term goal.

“The most experienced human ship captains will have the least trust in any single navigational device but will rather continuously cross reference between them. Similarly, our autonomous functionality will not be overly reliant on a single data source but combine and verify data from multiple sensors.

“Having gathered many gigabytes of data during our initial August field campaign, then again in October with more days planned in December, we are applying the results to train and test our data-fusing algorithms. A follow-up seagoing test will then verify their performance in practice.”

The Maritime AI-NAV team plans to employ a variety of sensor types, including satellite navigation receivers – also utilizing of Europe’s Galileo system — monocular and stereo cameras, standard radar, “laser radar” lidar and an array of microphones, along with “Automatic Identification System” radio signals. These AIS signals transmit position, size and routing information of all vessels above a certain class, as well as fixed infrastructure such as oil rigs or wind turbines.

“Satellite navigation lets the ship know where it is in the sea, while the other sensors let it know what is around it, which is essential for identifying and avoiding any obstacles,” Thombre said. “The different data sources operate across a variety of ranges — so radar and AIS provide longer range detection out to the horizon, while cameras and lidars come into their own at shorter distances. Plus we had a trio of microphones aboard the Megastar, determining the angle of arrival of sound from other ships. The challenge now is to fully integrate all these sources using machine learning, to build up a holistic picture.”

Maritime AI-NAV is supported through ESA’s Navigation Innovation and Support Programme, working with European industry and academia to develop innovative navigation technology.

FGI is joined in the Maritime AI-NAV consortium by Aalto University’s Sensor Informatics and Medical Technology group and maritime IT startup Fleetrange.

February 16, 2021
Why geospatial data needs artificial intelligence

By San Gunawardana, Guest Author

Advances in geospatial technology have opened up many new possibilities in areas such as national security, urban planning and emergency preparedness. When I was embedded with the U.S. Army as a scientist in Afghanistan, I got to experience firsthand the exceptional value of 3D data. The military used nation-scale imagery and lidar to generate 3D maps that then informed their safety-critical operations. However, since lidar—like most three-dimensional unstructured data—contains incredible complexity and detail, it was painfully slow to analyze manually.

As a result, the impact of this technology was severely restricted by speed and cost due to the significant manual effort required to extract actionable insights. As we looked to the future, where lidar would become commonplace in consumer electronics and automobiles, it became clear that there was an opportunity to combine computer vision/AI with large-scale cloud computing to rapidly and automatically generate actionable insights from 3D data.

Screenshot: Enview

After returning from Afghanistan, I reconnected with Krassimir Piperkov, a former colleague from ICON Aircraft, and fellow Stanford alum, to launch Enview. Our objective was to automate 3D geospatial analytics and create a living 3D model of the world to help organizations to protect their critical infrastructure and communities.

Powering geospatial data with AI can take the limits off 3D data analytics, prevent threats from becoming incidents, and protect critical infrastructure. What used to take days or months to process can now be done in minutes, enabling analysts, operators, and decision-makers across the public sector to make timely and accurate decisions. By enhancing our understanding of the physical world, this technology empowers us to tackle pressing challenges like wildfire prevention, humanitarian assistance, disaster response, and more.

Let’s take a look at how AI-powered 3D modeling is being put to use.

Digital twins

A living 3D model of the world, or a digital twin, can be used for many purposes. Enview’s software fuses many different data sets together to create digital twins that are global in scale but have high-resolution to enable local decision-making. These digital twins include 3D terrain, vegetation, buildings, and infrastructure such as power lines, roads, and water works. Enview also fuses real-time and forecasted conditions, such as wind, temperature, humidity, traffic, and IoT (internet of things).

This sort of rich representation of the physical world is an incredibly complex big data challenge. Data comes from radically different sensor modalities, with different resolutions, formats, time-domains, and accuracy. AI plays a critical role in automating the fusion of these datasets, by helping to intelligently align and then fuse them into a cohesive entity. 3D geospatial data is particularly challenging, as it is unstructured data, which requires a new generation of deep learning frameworks whose convolutional kernels are specifically developed from the ground up to work on unstructured data. Further, the datasets are massive in scale. A square-mile of 3D lidar data can have hundreds of millions of points; the magnitude of the data easily passes the petabyte scale when one considers applications that span nation-scale areas. In order to process this volume of data, modern geospatial AI architectures must be containerized and dynamically deployable across cloud compute resources to generate timely insights.

AI is essential to help human experts to extract meaningful insight from this overabundance of data. The application of automated workflows allows experts to look at larger areas, with more speed and higher frequencies. This machine-assisted cognition draws upon the respective strengths of people and computers to do what neither could do on their own.

Humanitarian aid and disaster relief

3D models can be built to monitor hurricane hotspots, such as the Gulf Coast, before major storms strike. By layering in real-time weather information such as rainfall, winds, and flooding, these models can help with planning, emergency response, and relief efforts.

This data also provides life-saving insight that can assess damage to buildings, transportation, and downed power lines, in addition to determining where to send medical and relief supplies, and how to best get them there. 3D data can help to lessen the impact of future weather events by updating the baseline understanding of how storms impact coastal communities so they can plan for the future.

Screenshot: Enview

Infrastructure protection

Inadequate clearances between vegetation and power lines can result in wildfires and unplanned power outages. Many federal, state, and local regulations are in place to mandate clearances, and power line operators monitor their networks continuously to ensure that they abide by these regulations and prevent incidents and outages. However, doing so by walking or flying the lines and judging distances with the human eye is challenging and inaccurate.

The ability to identify the exact location and clearances of high-risk vegetation early, and at scale, lets operators identify, prioritize, and address problem areas proactively. Lidar-driven programs have helped with risk-reduction, but are constrained by the massive levels of manual data manipulation required to derive insights from this 3D data. The automation of 3D geospatial analytics through AI, machine vision, and parallel computing enables the accurate and rapid identification of at-risk areas, protecting critical infrastructure and communities.

Screenshot: Enview

Fighting wildfires

Devastating wildfires resulting in the loss of life and property have become commonplace in the western U.S. and other parts of the world. The tools and methods previously relied on to keep communities and infrastructure safe are now struggling to keep up with this increased threat.

Geospatial information, including 3D data, provides a digital view of the physical world and, when paired with AI, gives stakeholders the informational edge they need to minimize wildfire damage, injuries, and deaths. This technology can be used to automatically build and update real-time, high-resolution wildfire risk maps that give firefighters and communities more notice when threats are imminent, and provide firefighters with real-time situational awareness when they’re fighting the blazes.

Change detection

According to the Pipeline and Hazardous Materials Safety Administration (PHSMA), third-party excavations are one of the leading causes of pipeline incidents in the U.S. These incidents can lead to service disruptions, expensive repairs, and sometimes serious injuries or deaths.

Detecting signs of excavation or earth movement via aerial patrolling is challenging and costly, while resource limitations make it difficult for pipeline operators to continuously monitor remote areas such as farms. AI-powered 3D maps can be used to monitor topography and accurately detect changes that threaten pipelines in real time.

3D data provides remarkable value when it comes to decision-making as it relates to many different applications—from military defense to protecting neighborhoods from wildfires. However, its success hinges on one thing: speed. The ability to process 3D geospatial data rapidly, and at scale, is made possible through advances in AI and cloud computing. In the future, we can expect to see more exciting and innovative use cases for AI-powered geospatial technology.

San Gunawardana is co-founder and CEO of Enview, a geospatial analytics company. After finishing a Ph.D. in aerospace engineering at Stanford, Gunawardana went to Afghanistan, where he combined data analytics and remote sensing to detect threats and prevent incidents. He is excited to apply those insights to help the energy sector solve problems. He has done computer vision at NASA, built imaging satellites with the Air Force, and was an early employee at ICON Aircraft.

January 27, 2021
Daewoo E&C partners with SPH Engineering for AI platform

SPH Engineering has partnered with Daewoo Engineering and Construction (E&C). Through the partnership, SPH will support Daewoo’s data management projects through its Atlas artificial intelligence (AI) platform, which enables aerial imagery storage, map creation, change tracking, object detection and territory segmentation.

Photogrammetry data is expected to become one of the key components for storage and processing, SPH added.

According to the companies, Atlas will enable Daewoo Engineering and Construction to set up an online archive of drone imagery and photogrammetry products, track changes and generate reports, automate object detection and measure the identified objects of interest. The platform also will increase data availability for participants of construction workflow.

“Atlas can be definitely used in various fields, but it will be a groundbreaking platform, especially in the field of construction survey,” said Geunmok Song (Alex), digital construction team manager at Daewoo Engineering and Construction.

“When we introduced Atlas back in spring, first of all we wanted to support our existing UgCS customers with an easy-to-use AI tool to store and process data collected with our software integrated to a UAV,” said Alexei Yankelevich, R&D director at SPH Engineering. “We are proud that Daewoo Engineering and Construction, the representative of Korea, has opted for our solution.”

November 18, 2020
New imaging method uses time to create pictures

Alex Turpin (Photo: University of Glasgow)

A new method of imaging that harnesses artificial intelligence to turn time into visions of 3D space could help cars, mobile devices and health monitors develop 360-degree awareness.

Photos and videos are usually produced by capturing photons with digital sensors. 3D images can be generated either by positioning two or more cameras around the subject to photograph it from multiple angles, or by using streams of photons to scan the scene and reconstruct it in three dimensions. Either way, an image is only built if spatial information of the scene is gathered.

Now, researchers based in the United Kingdom, Italy and the Netherlands describe how they have found an entirely new way to make animated 3D images — by capturing temporal information about photons instead of their spatial coordinates. The team’s paper, “Spatial images from temporal data,” was published in Optica.

Their process begins with a simple, inexpensive single-point detector tuned to act as a kind of stopwatch for photons. Unlike cameras, which measure the spatial distribution of color and intensity, the detector only records how long it takes the photons produced by the split-second flash of a pulse of laser light to bounce off each object in any given scene and reach the sensor. The farther away an object is, the longer it will take each reflected photon to reach the sensor.

The information about the timings of each photon reflected in the scene — temporal data — is collected in a simple histogram. Those graphs are then turned into a 3D image using a sophisticated neural network algorithm. The researchers “trained” the algorithm by showing it thousands of conventional photos of the team moving and carrying objects around the lab, alongside temporal data captured by the single-point detector at the same time. Eventually, the network learned enough about how the temporal data corresponded with the photos that it was capable of creating highly accurate images from the temporal data alone.

In the proof-of-principle experiments, the team managed to construct moving images at about 10 frames per second from the temporal data, although the hardware and algorithm used has the potential to produce thousands of images per second.

Alex Turpin, a Lord Kelvin Adam Smith Fellow in Data Science at the University of Glasgow’s School of Computing Science, led the university research team with Prof. Daniele Faccio and support from colleagues at the Polytechnic University of Milan and Delft University of Technology.

“Cameras in our cellphones form an image by using millions of pixels,” explained Turpin. “Creating images with a single pixel alone is impossible if we only consider spatial information, as a single-point detector has none. However, such a detector can still provide valuable information about time. What we’ve managed to do is find a new way to turn one-dimensional data — a simple measurement of time — into a moving image that represents the three dimensions of space in any given scene.”

After data collection, 3D images are retrieved from the temporal histograms. (Image: University of Glasgow)

The approach is capable of decoupling light altogether from the image-capture process, and the paper discusses how the team managed to use radar waves for the same purpose. “We’re confident that the method can be adapted to any system which is capable of probing a scene with short pulses and precisely measuring the return ‘echo.’”

Right now, the neural net’s ability to create images is limited to what it has been trained to pick out from the temporal data of scenes created by the researchers. But with further training and by using more advanced algorithms, it could learn to visualize a range of scenes, widening its potential applications in real-world situations.

“The single-point detectors that collect the temporal data are small, light and inexpensive, which means they could be easily added to existing systems like the cameras in autonomous vehicles to increase the accuracy and speed of their pathfinding,” Turpin said. “Alternatively, they could augment existing sensors in mobile devices like the Google Pixel 4, which already has a simple gesture-recognition system based on radar technology. Future generations of our technology might even be used to monitor the rise and fall of a patient’s chest in a hospital to alert staff to changes in their breathing, or to keep track of their movements to ensure their safety in a data-compliant way.”

Next, the team will work on a self-contained, portable system-in-a-box as well as examining options for furthering research with input from commercial partners. The research was funded by the Royal Academy of Engineering, the Alexander von Humboldt Stiftung, the Engineering and Physical Sciences Research Council (ESPRC) and Amazon.

Citation. A. Turpin, G. Musarra, V. Kapitany, F. Tonolini, A. Lyons, I. Starshynov, F. Villa, E. Conca, F. Fioranelli, R. Murray-Smith, and D. Faccio, “Spatial images from temporal data,” Optica 7, 900-905 (2020), https://doi.org/10.1364/OPTICA.392465.

October 22, 2020
Qualcomm launches 5G, AI-enabled robotics platform

The Qualcomm Robotics RB5 Development Kit (Photo: Qualcomm Technologies)

Qualcomm Technologies released the Qualcomm Robotics RB5 platform. The RB5, comprised of hardware, software and development tools, is designed for the consumer, enterprise, defense, industrial and professional service sectors.

According to the company, the platform’s Qualcomm QRB5165 processor offers a heterogeneous computing architecture, coupled with the fifth-generation Qualcomm AI Engine that delivers 15 tera operations per second of artificial intelligence (AI) performance for running complex AI and deep learning workloads. The processor also offers incredible machine learning inferencing at the edge under restricted power budgets using the new Qualcomm Hexagon Tensor Accelerator.

Technical features of the RB5 include heterogeneous computing capabilities, 5th generation Qualcomm AI engine, advanced imaging capability, security support and connectivity. Qualcomm’s Spectra 480 Image Signal Processor (ISP) captures fast, professional-quality photos and videos, and can process two gigapixels per second, the company said.

In addition, seven concurrent cameras facilitate simultaneous localization and mapping (SLAM), object detection and classification, autonomous navigation and path planning to perform tasks in indoor and outdoor settings.

With the Qualcomm Robotics RB5 platform and the Qualcomm QRB5165 processor, Qualcomm enables various design offerings including off-the-shelf system-on-module solutions and flexible chip-on-board designs, the company said. The solution is available in multiple options, including commercial and industrial-grade temperature ranges and an option for extended lifecycle until 2029.

“With the Qualcomm Robotics RB5 platform, Qualcomm Technologies will help accelerate growth in a wide array of robotics segments such as autonomous mobile robots, delivery, inspection, inventory, industrial, collaborative robots and unmanned aerial vehicles, enabling Industry 4.0 robotics use cases, and laying the foundation for the UAV Traffic Management space,” said Dev Singh, senior director, business development and head of autonomous robotics, drones and intelligent machines at Qualcomm.

Qualcomm also has entered into a strategic collaboration with TDK to further enhance the capabilities of the Qualcomm Robotics RB5 platform. Through the partnership, TDK added its latest sensor technologies for enhanced robotics applications as part of the Qualcomm Robotics RB5 platform.

The Qualcomm Robotics RB5 Development Kit

In addition, Qualcomm Robotics RB5 Development Kit ensure developers have the customization and flexibility they need to make their visions a commercial reality.

According to Qualcomm, the kit allows developers to have flexible software capabilities, with the platform offering support for Linux, Ubuntu and Robot Operating System 2.0, as well as pre-integrated drivers for various cameras, sensors and 5G connectivity. It also provides support for OpenCL, OpenGLES and OpenCV.

It also includes support for the Intel RealSense Depth Camera D435i and Panasonic TOF Camera to provide depth-sensing capabilities. TDK’s six-axis ICM-42688-P IMU, ICP-10111 barometric pressure and T5818 Digital bottom port microphone are integrated into the kit, as well.

June 18, 2020
ADVA tackles GNSS jamming and spoofing with AI solution

ADVA has launched a centralized GNSS monitoring and assurance tool that uses artificial intelligence (AI) and machine learning (ML) for comprehensive predictive maintenance.

The new customer-owned tool enables users to collect and analyze huge amounts of information from across the network to remotely identify issues and protect networks from GNSS vulnerabilities, including jamming and spoofing attacks.

It also helps to identify GNSS obstruction issues, detect blind/poor spots that appear over time, and enable optimal antenna positioning.

Built into ADVA’s Ensemble Controller network management suite with Sync Director, the solution enables customers to detect potential problems in advance, maintain the highest quality of network synchronization and significantly reduce opex. By complementing today’s limited distributed approach to GNSS assurance with a centralized-global system, it offers a major boost to critical infrastructure dependent on satellite-based timing.

“What we’re offering is a way for network operators to see the bigger GNSS picture. Using AI and ML to analyze the entire synchronization network, our centralized GNSS monitoring and assurance solution will be key in the fight against GNSS cyber issues, such as jamming and spoofing attacks,” said Gil Biran, general manager, Oscilloquartz, ADVA.

“This new technology provides the power to proactively tackle issues that jeopardize vital services,” Biran said. “Harnessing the capabilities of our synchronization devices to identify spoofing problems, it intelligently mines a wealth of data and gives network operators the precise info they need in a highly accessible way. By using long-term heat maps and enormous amounts of data from a wide range of GNSS receiver sources, our solution identifies patterns and preempts issues. It alerts maintenance teams to obstructions or jamming conditions so that countermeasures can be put in place well before services are affected.”

As part of the network infrastructure, ADVA’s centralized GNSS assurance and monitoring solution enables a network-wide view of GNSS receiver health. Requiring no additional hardware or site visits, it remotely delivers detailed analysis, automatically detecting abnormal patterns with a patent-pending algorithm.

Utilizing AI and ML, it alerts maintenance teams to potential GNSS service degradation and safeguards against spoofed signals. Network operators receive updates through a user-friendly GUI as well as regular reports tailored to individual criteria.

As a component of ADVA’s comprehensive Ensemble Controller suite, the new technology makes synchronization monitoring and assurance an integral part of overall network management and control. For network operators, having a single system to track inventory simplifies operations and helps bolster network security.

“GNSS is the fundamental source of network time, phase and frequency generation across so many of today’s industries. From IT to telecommunications, from energy to finance, the reliability of satellite-based timing is crucial and the cost of interference is huge. This latest launch is a key part of our ongoing mission to remove the risk of GNSS vulnerabilities,” said Nir Laufer, senior director, product line management, Oscilloquartz, ADVA.

“The new solution joins our multi-band, multi-constellation GNSS receiver technology — which overcomes ionospheric delay variation — as well as our range of grandmaster clocks with network-based timing and outstanding holdover capabilities,” Laufer said. “Combined with our highly stable cesium clock technology, these create our ePRTC solutions for ultimate GNSS backup. With our comprehensive portfolio, all industry verticals are guaranteed accurate, cost-effective and highly resilient timing.”

February 12, 2020
Drones and imagery: Utilities turn to artificial intelligence

How AI and machine learning algorithms redefine the way utility companies manage their infrastructure

By Jaro Uljanovs, Lead AI Developer and Data Scientist, Sharper Shape

Artificial intelligence (AI) boasts a wide range of potential applications, across nearly every industry imaginable — healthcare, automotive, retail, even fast food. But it’s the utility industry where AI and machine learning (ML) are beginning to demonstrate some of their most impactful effects on many aspects of the business. Power companies are increasingly leaning on AI to improve their electricity delivery and prevent potential wildfires, and AI is actually enhancing, rather than eliminating, human jobs.

From data collection and analysis to their presentation of actionable insights, AI and ML algorithms are quickly redefining how utility companies manage their electric infrastructure.

Consolidating and classifying data

Utility companies oversee massive infrastructure networks, comprising poles, conductors, substations and transmission and distribution lines that span thousands of miles. The vegetation surrounding this key infrastructure must also be monitored, as it presents a danger of fire or outage.

Taking a comprehensive snapshot of these assets means utilizing a variety of different sensors for network inspections. These sensors include lidar, color (RGB), hyperspectral and thermal imagery.

This allows the system to capture everything — from vegetation proximity, to infrastructure assets, to individual components (such as insulators on poles) and their operational integrity, to hot spots indicating potential fire risks.

That’s a lot of data to capture, catalog and process. And there are a lot of individual elements within that data — even in just one image — to pinpoint and classify, let alone do so accurately. Classifying billions of data points across all of those images is an impossibly time-consuming task to do manually.

Photo: shaunl/E+/Getty Images

AI and ML tools can accomplish that same work — scanning thousands of images collected across thousands of miles of utility infrastructure — in seconds. Lidar point cloud segmentation can detect conductors (quite a difficult component-type to segment) with an accuracy of over 90%, while hyperspectral image segmentation can identify vegetation species with an accuracy of up to 99%.

More than that, when paired with drone sensors, these algorithms can also improve the upfront collection of images and data. AI and ML tools help to adjust sensor positioning in real time, in the event a signal is lost or the drone veers slightly away from its inspection flight path.

By helping to readjust the sensors’ bearings while in flight, AI not only ensures more accurate data collection, but also that the flight doesn’t need to be done again or prematurely ended because of faulty data collection, saving time and money. AI pinpoints any faults in the sensors or the drone’s flight path while in the air, recalibrating as needed and identifying individual elements within the data as it comes through the sensor’s video feed.

Breaking down silos to create a holistic data approach

Key to all of this is eliminating the silos that tend to naturally build up between different data segments. In the utility inspection space, asset management, vegetation management, different sensors and so on all produce their own disparate, walled-off sets of data.

When data is kept siloed like this, it becomes unnecessarily difficult if not impossible for teams to derive companywide insights or conclusions from the information being collected. And what good is all that data if it can’t be used to check against itself and enhance other sets of data?

The northwest fire line of the wildfire that devastated Santa Rosa, California, taken by satellite Oct. 10. (Satellite image ©2017 DigitalGlobe)

Good data management can’t exist in a piecemeal approach. It needs to be holistic, and AI provides the impetus to make that happen. AI provides a central resource for pooling all these data sources together, making it easier to cross-analyze for potential problems — like wildfire-prone vegetation or damaged components. When these issues are collected in one system, it becomes much easier to identify faults and resolve them — and do so far faster than it would be to manually sift through countless images of poles or vegetation maps.

And for all the stereotypical concerns about AI eliminating work for human beings, at utility companies AI actually enhances the role that people have to play in the network inspection process. Because the AI is what analyzes the data, it’s not something that is dependent on the potentially biased expertise of a professional human inspector, nor is it prone to fatigue and the anomalous results that can come from that. But at the same time, AI can’t do everything itself. It’s a tool for presenting clearer, more accurate and more actionable information for the people to then act on with their own judgment.

There’s a lot of easy-to-make assumptions, both good and bad, about AI. But at the end of the day, what AI really means for the utility industry is a more efficient and effective tool for providing the right information about a power company’s infrastructure — its transmission and distributions lines, its poles, and its nearby vegetation — into the hands of its key decision makers.

December 19, 2019