Dr. Martin, tell us a little bit about your background.
My career spans nearly 35 years in aerospace, with more than 20 years as a military aerospace engineering officer, in both the Australian airforce and army, on a variety of fixed and rotary-wing platforms, manned and unmanned. My last role in the Australian Army was as the Training Systems Manager for the Armed Reconnaissance Helicopter, overseeing the acquisition of more than $250 worth of simulators and training assets. My unmanned experience includes work on Scan Eagle, Shadow, Heron, and Triton.
I hold a PhD in Machine Learning and Pattern Classification and am an Adjunct Professor at Queensland University of Technology (QUT). For purposes of this piece, I”M currently the Project Director for a Nova-led UTM consortium in Singapore. Nova is a technology and professional services company with its genesis in military flight test and evaluation, and the consortium partners include a mix of Australian, Singaporean and US companies including M1, Amazon, OneSky, Rohde and Schwarz, and Scout Aerial.
Detect-and-avoid (DAA) continues to be a challenge globally and one of several hurdles to fully implementing UTM. For example, in the U.S., NASA has been exploring DAA for years. What progress are you seeing in this area, in Singapore and other parts of the world?
While we are not addressing DAA as part of the Singapore project, I have done some research in Australia on that front, and am continuing to supervise some engineering students at QUT, with a colleague, Dr. Aaron Mcfadyen. We are looking at light-weight medium -wave infrared systems for low size, weight and power (SWAP) platforms.
The fielding of DAA for collision avoidance continues to pose challenges. I think the answer lies in acknowledging that you don”T need the gold plated solution that has historically been imposed on large military platforms, particularly if you can show that for low traffic environments, you can still meet regulatory expectations for Target Levels of Safety with DAA system that perform the detect part of the ?detect-decide-move? feedback loop at say 70% rather than 99%.
What I will say is that concession needs to be supplemented by quantitative evidence demonstrating what the actual traffic is, through airspace characterization or a qualitative decomposition accepted by regulatory bodies, in conjunction with containment and an acceptable response time horizon. For example, a 60 kt Scan Eagle-like platform and a Cessna at 120 kt still have a closure speed of 3kts/min (approx 100 m/s), which means you are probably going to want to detect an intruder 20-30 seconds away which equates to a detection distance of at least 2-3km. If you then superimpose usable miss/false alarm rate setups, factoring in day to day air and ground background clutter, this is extremely challenging for low SWAP systems. That’s why in my other collaborative research side-hustles, I”M focused on solid-state mid-wave infrared (MWIR) solutions
Ultimately, the difficulty of the DAA tasks is what motivated me to pivot my focus to UTM, probably via an inadvertent push from PK from NASA
In 2015, as part of some electro-optical DAA research I was doing, I came across a paper from the U.S. Office of Naval Research, which surmised that rather than waste another 10 years of time and a $1 Billion in funding trying to deal with the challenge of electronically invisible VFR aircraft, just make everything electronically conspicuous. At the time, this was almost a blasphemy. Simultaneously, Parimal Kopardekar (PK) was in the early stages of evangelizing UTM and provided a brief in Singapore. I was instantly fascinated. PK extended an invite to NASA and I jumped at the chance. I have been a zealot for UTM ever since!
Recently U.S. UAS regulations were updated to use the term ‘sense and avoid.” What have you seen in Singapore in regulation, that addresses this?
I think Singapore is keeping their eyes on what the U.S. and Europe are doing and picking the best bits of both. They have a pretty good system dealing with Ground Risk, but by their own admission, the Air Risk element is still in development. In saying that, as an island state, with significant Air Force oversight on their airborne risk environment, and very little recreational visual flight rules (VFR) traffic, the risk of uncooperative intruders is insignificant. (See Swiss example here: https://www.bazl.admin.ch/dam/bazl/en/dokumente/Gut_zu_wissen/Drohnen_und_Flugmodelle/uspace_conops.pdf.download.pdf/Swiss%20U-Space%20ConOps.pdf). So by default, they are probably going to achieve a system where everyone is electronically conspicuous. (See Singapore regulations here: https://www.caas.gov.sg/docs/default-source/default-document-library/ac-anr101-2-2-bvlos-operations-for-ua_301219.pdf)
You mentioned the importance of a comprehensive air picture. That is something that the U.S. and all countries are trying to achieve as part of UTM. What are your thoughts on the means to achieve that?
This is a tricky question. Intuitively a comprehensive air picture would incorporate the surveillance data captured by the Airspace Navigation Service Provider (ANSP) for manned aircraft, as well as the data for unmanned aircraft vehicles (UAVs) in low-level airspace. In the U.S. you call it the Flight Information Management System (FIMS). In parts of Europe, they make reference to a Common Information Service or Function (CIS) which subsumes the FIMS, but also consolidates a number of functions where there is common information, such as drone registries, local and national geo-zone repositories, pilot information. There is a bit of contention over the demarcation for where industry, ANSPs and regulators should manage that information and the derived services.
These systems help provide a better picture of where everything is flying. But that knowledge is useless unless we use that information to keep everything safely apart. For a little while, ?blue sky? will be your friend, but if the VC’s actually want to achieve the scale necessary to get a Return on Investment, then someone has to solve the problem of optimizing airspace capacity management, and ensuring that package-delivering UAVs operating in proximity to Urban Air Mobility (UAM) vehicles, in low-level airspace are safely separated from each other, and potentially, other manned platforms.
Fundamental to this will be understanding of how accurately you can measure where each platform is, how long it takes to intervene from an ATC/(automated unmanned traffic controller or UTC) perspective, and when you can relax your expectations. For instance, in a metropolitan environment like Singapore, the Global Navigation Satellite System (GNSS) solution and communications connectivity can be compromised because of multipath effects (e.g., interference from buildings). We’ve validated this in our research. During flights, this interference obscures the line of sight to many satellites, ultimately diluting the precision of the navigation solution and potentially compromising safety, because of position uncertainty.
Worse, if you are also Beyond Radio Line Of Sight (BRLOS), and otherwise reliant on either 4/5G or SATCOM, your ability to intervene can also be negatively affected by CNPC latency in two ways. First, your ability to command the aircraft to do what you want in the event you are directed to change course, and second, to act as a conduit for surveillance or conformance monitoring. But this is not all, both the operator and the UTM Service Supplier need to be aware of the Total Systems Error (the difference between the actual flight path of an aircraft and the assigned flight path, equal to the sum of the navigation system error, Flight Technical Error (FTE) and Path Definition Error(PDE)) in tandem with appropriate alert times.
Consensus on this is only starting to emerge within ASTM and the role it plays in strategic and tactical deconfliction, but we have been working on this for nearly two years as part of our project scope
It is also worth emphasizing that for the manned environment, the amount of PDE is typically low, but for UTM, managing the proximity of UAVs to building is going to require accurate building data for both route design and obstacle avoidance. This is not commonly available information in most countries.
Are UAVs in Singapore operating within set distances from launch platforms with pre-positioned way points, providing right and left limits of an operations box, and allowing freedom of movement within it?
Yes. We have designed a system called MARS: MultiAgent Routing and Scheduling. It allows users to input their desired route, from either A to B, or to operate within an area. A 4D volume, or series of volumes is created, and then a route is planned and scheduled amidst a fleet of operations for our UAS service suppliers (USS). We have now made our system ASTM-compliant, such that the Intended Operational Flight Volumes can be submitted to the Discovery and Synchronisation Service (DSS) and we can then negotiate any conflicts with other USSs in the Identification Service Area (ISA), which under the ASTM standard is defined as a 4-D volume corresponding to an area where a Net-RID Service Provider has one or more UAS operating.
At present, ASTM has not settled on how conflicts will be prioritized and negotiated, but we are incorporating our own rules based on communication, navigation, and surveillance (CNS) and Total Systems Errors we have determined based on data captured during flight trials. So users will have the perception of complete freedom, but some constraints may be embedded in our routing network to optimize flow.
Are you using Remote identification (RID) capabilities or testing them?
Yes, we are using RID, in compliance with the ATSM standard. Singapore has also adopted that.
Until the ASTM standard was released, we were using our own proprietary protocol but now we have made that system compatible with the Network RID system. We intend to test that out in August and September in Singapore, and probably as part of Strategic Deconfliction trials with Swiss U-Space Implementation (SUSI) platform, created by the Swiss Federal Office of Civil Aviation (FOCA). U-Space consists of a set of decentralized services created to integrate drones in the airspace and to enable them to operate together with manned aircraft, in a transparent way.
One thing to note in regard to implementation. Our telemetry data goes directly from the aircraft to our Amazon Web Services (AWS) telemetry server via Long-Term Evolution (LTE), in addition to being fed back via an radio frequency (RF) link to our Ground Control Station (GCS). Many other systems have the telemetry go from the UAV back to the GCS via RF, then into the network, potentially introducing additional latency and a single point of failure. We are also looking at incorporating dual sims with separate providers for extra redundancy.
The proposed U.S. RID rule prohibits use of ADS-B for small UAS. Does Singapore have that same restriction?
I can”T speak to what the Singapore regulator will do, but it would surprise me if the deviated with the emerging consensus, and I don”T believe ADS-B will get support from regulators for a variety of technical reasons. We also do not believe it could be implemented in a timely manner, particularly the cost and latency that would occur due to Aviation Navigation Service Providers (ANSPs) having to submit a Contract Change Proposal to modify their existing systems.
How do you communicate with and between aircraft?
We communicate with the aircraft via two means. First, we communicate directly from our GCS over 4G/5G networks to the UAV using an onboard puck with a subscriber identification module (SIM) card. Second, we use direct RF, which is obviously less effective as platforms move BRLOS.
In terms of inter-aircraft comms, we have not implemented anything to date, except to rely on Remote ID and the emerging federated implementation of UTM which would connect to a FIMS/CIF setup. This ostensibly could be used in a similar manner to the way ADSB-in and ADSB-rebroadcast is deployed.
How do you ensure the reliability of required communications, given that GPS is difficult in urban areas because of tall buildings and related satellite interference?
We have done a lot of work capturing Dilution of Precision around buildings, as a function of altitude, and in a lot of places. The results are not pretty.
We haven”T done a lot of work on alternatives, but perhaps some hybrid of multilateration techniques in conjunction with kalman filters might emerge. (Multilateration is a technology for determining the position of an emitter (from a platform transponder or transmission) by measuring the Time Difference of Arrival (TDOA) of the signal at several known observation points, and with a priori knowledge on the speed of propagation, you can triangulate position via hyperbolic positioning). Resolving a platform with RF depends on frequency and power. Many of the frequencies that telecommunications companies (telcos) transmit on are not going to give you the accuracy you need if you multilaterate via telco base-stations. 5G at higher frequencies might work, but it will require more base-stations.
Are you taking into account a possible dedicated radio frequency for public safety, and used for priority permissions in emergencies and disasters? How?
We have done a little bit of work on this front, first in Australia with Telstra, the country’s largest mobile network, using Internet Protocol (IP) Addressable channels, called Telstra LANES. It is perfect for emergency services because you have a pipe where nobody else is cluttering it up with network traffic. We did something similar in Singapore with their mobile network called M1.
I imagine this is probably the way it will go in other countries. I also note that in Singapore, there has been some experimentation using a slice of bandwidth made available from a defunct radio station, which isn”T a bad idea for supporting low data requirements such as safety critical control and basic telemetry.
But, I need to emphasise LANES or its equivalent, can ensure your signal isn”T corrupted because of network traffic: it’s not immune to signal propagation corruption via buildings and terrain.
How might you provide relevant information to ATC or USS but not so much information as to overwhelm the humans receiving it?
We have gone down the path of exploring this, because many of Nova’s staff are military test pilots and engineers, where Human Factors (HF), the study of the interrelationship between humans and the tools and equipment they use and the environment in which they work, are a significant consideration.
In our last trials, we explored a bunch of HF elements, particularly around the graphic-user interface (GUI) and how useful it is during normal operations and other contingencies. We paid particular attention to the Bedford Workload Rating scheme, which looks at the operator’s ability to perform his primary and secondary task, or whether he had very little capacity due to the difficulty of the task or poor interface. Of course, dealing with contingencies is often a combination of things, with experience also being a factor in overcoming poor design. It’s my belief that the USS HMI will start to attract greater scrutiny in the fullness of time, as the amount of traffic and accidents increase, in a similar way that aircraft cockpit’s HMI have improved over the decades, but also needed to reconsider the impacts of new technological products like glass cockpits, Helmet Mounted Displays and Night Vision Goggles.
Part two of the article coming next week!
*The views and opinions in this article are those of the author and do not reflect those the DOD, do not constitute endorsement of any organization mentioned herein and are not intended to influence the action of federal agencies or their employees.
Dawn M.K. Zoldi (Colonel, USAF, Retired) is a licensed attorney and a 25-year Air Force veteran. She is an internationally recognized expert on unmanned aircraft system law and policy, a recipient of the Woman to Watch in UAS (Leadership) Award 2019, and the CEO of P3 Tech Consulting LLC.