Clear Prop! | January Papers (1/2)
eVTOL Crash Location Analysis, Quantifiable Public Acceptance, Collision Avoidance Algorithms & more
Welcome to the first ever edition of the Clear Prop! newsletter! My aspiration is to make cutting-edge research accessible to all: entrepreneurs, engineers, investors, alike.
I’ll be starting by summarizing 5 research papers that I’ve found exciting in AAM. Unlike other newsletters that focus on the hot events that have just arrived from the combustion chamber, I will begin in January 2022, successively making my way through the months. Academic research is much slower-paced with intellectual & commercial impact measured in decades, so there is no rush.
Before we dive in, if you haven’t seen the new Top Gun movie, you are either living on Mars with no internet or dislike airplanes (!) Anyways, Tom Cruise took James Corden up into the sky again. This time, it’s not skydiving but they flew with epic aircraft such as the North American P-51 Mustang (Cruise’s legendary WWII airplane) and the Aero L-39 Albatros (Warsaw Pact bird, the most popular jet for training). Enjoy.
#1: Research on The Crash Location and Speed Distribution of Low Altitude Fixed-Wing Aircraft
In this paper from Inha University (Republic of Korea), the authors analyze the potential crash locations and impact speeds for a Hyundai S-A1 eVTOL losing all engines & control at 500m AGL, at a cruising speed of 240 km/h. They ran 10k+ simulations, running through various initial conditions for different aileron & elevator angles using a Monte Carlo approach. The simulation resulted in a probability distribution of crash locations and impact speeds.
Key takeaways:
The crash location can be as far as 9.1 km. However, 95% of all crashes will be located somewhere 400 m ahead.
The impact speed can be as high as 360 km/h. However, most impact speeds average around 200-220 km/h.
Even a small variation in control surface angles can make a dramatic difference in crash location and impact speed.
The BFD: Korea is currently one of the hotspots for AAM. Home to Supernal (a Hyundai company) which is developing an eVTOL & mobility service, Korea is also seeing considerable government support in establishing a UAM ecosystem. Western firms such as Vertical Aerospace and Overair have already announced their partnerships in the country. Thus, academic research on operational safety coming out of Korean universities makes sense for national AAM roll-out both from a commercial and technical perspective.
#2: A Fast Markov Decision Process-Based Algorithm for Collision Avoidance in Urban Air Mobility
Researchers from Iowa State and George Washington Universities propose a collision avoidance algorithm for drones & eVTOLs based on Markov Decision Processes (MDPs). MDPs are a key building block of reinforcement learning (think AlphaGo) and are fundamental to the forthcoming ACAS-X, an advanced advisory system for pilots for impending 1-on-1 collisions. Called FastMDP, the authors compare their algorithm to existing collision avoidance approaches out there including Monte Carlo Tree Search (MCTS) and Optimal Reciprocal Collision Avoidance (ORCA). Comparisons are made across the following metrics: 1) throughput (think traffic flow), 2) computational performance (i.e. how fast the algorithm runs), 3) near mid-air collisions and 4) loss of separation.
Key takeaways:
FastMDP executes ~5x faster than MCTS and ORCA.
FastMDP exceeds ORCA’s collision avoidance performance. It closely trails behind MCTS while sacrificing only a modest amount of vehicle throughput.
FastMDP is only applicable to a certain subset of MDPs and cannot currently solve a general MDP problem.
The BFD: Ensuring safety in the skies where drones & eVTOLs will operate is key. However, with 100s or even thousands of aircraft in the air doing medical deliveries and transporting commuters over a city, a centralized traffic management system will need additional “fail-safe”redundancies such as decentralized, on-vehicle collision avoidance systems. Implementing a FastMDP-like algorithm may be one way of doing it. Nonetheless, the introduction of approaches that are fundamental to autonomy (e.g. MDPs in reinforcement learning) will aid in the gradual transition to non-deterministic (think AI) systems within the legacy aviation community.
#3: Trajectory Analysis in a Lane-Based UAS Traffic Management System
In this paper from the University of Utah, the researchers propose an approach to detect rogue aircraft within a flight lane. Much like how cars drive on highways, the initial roll-out of eVTOLs will likely have “skyways”. These pre-determined flight corridors (much like Victor & Jet Routes today) are aimed to constrain the environment for certain vehicles to maximize safety while reducing the workload on air traffic management systems. The authors develop a machine learning (ML) algorithm to classify planned flights from those that are rogue, based on simulated trajectories.
Key takeaways:
By extracting key features from example rogue flights such as hobbyist or criminal, the ML algorithm can classify 100% of the unplanned flights correctly.
Although not given experimental evidence, it is inferred that this is a much more efficient way of determining rogue flights in the airspace vs free-flowing traffic (i.e. without skyways) would require.
Further work is necessary to experimentally test the results in a real world setting.
The BFD: For the past few years, there has been intense debate on how to structure urban airspace. The advent of Unmanned Traffic Management (UTM) - a digitized & highly automated version of air traffic control - makes it possible for flights to be planned in a free-flowing, unconstrained manner. This 4-dimensional, skyway-free approach is the future for urban airspace to achieve maximum capacity with minimum delays. However, for this to happen in congested airspace, non-deterministic systems (think AI/ML, neural networks, reinforcement learning, etc.) will need to be accepted by aviation regulators. Until then, a skyway or 1-dimensional approach to flight routes makes sense. This is what NASA and the FAA also advocate, at least initially. Solutions to ensure the safety of such skyways will be key, as authors of this paper demonstrate in their approach.
#4: Life Cycle Engineering Modelling Framework for Batteries Powering Electric Aircrafts – the Contribution of eVTOLs Towards a More Sustainable Urban Mobility
In this paper from the Technical University of Braunschweig (Germany), researchers propose a framework for quantifying the true carbon footprint of eVTOL batteries. Based on integrated computational life cycle engineering, the carbon footprint can be calculated based on vehicle hardware, operational, geographical and temporal factors. The authors simulated over 11k routes across 24 cities around the world, experimenting with different battery chemistries, under differing weather conditions, and the like.
Key takeaways:
The carbon footprint of eVTOLs are highly sensitive to operation in countries with a dirty electricity mix (e.g. coal) and hot weather
High load factors (i.e. more passengers in the vehicle) and longer routes are key to reducing the carbon footprint on a per-km, per-pax basis
Battery lifetime, along with the specific energy (i.e. how much energy is packed into the battery), has a sizeable impact on the carbon footprint
The BFD: The environment, in addition to vehicle/battery tech, plays a huge role in making electric aircraft sustainable across the life cycle. While most of the media & industry attention goes towards batteries, the AAM ecosystem will need to consider factors both upstream & downstream to the vehicle itself. eVTOL operators launching in a given market will need to pay attention to the precise energy mix at the grid while optimizing their routes, flight profiles and use cases to environmental conditions. Thus, the framework proposed by this paper provides a methodology to quantifiably and responsibly roll-out aerial mobility services.
#5: Implementing Mitigations for Improving Societal Acceptance of Urban Air Mobility
Researchers from the Technical University of Catalonia, HEMAV Foundation, and EUROCONTROL introduce a quantifiable framework for tackling public perception issues for drones & eVTOLs. The authors begin by assessing previous surveys done by the likes of McKinsey, EASA, Airbus, etc. and attempt to explain the large variability of results between these studies. Later on, a new methodology is introduced that categorizes various public concerns about aerial vehicles such as noise, safety, and privacy. Potential mitigation measures for each are established, then to be ranked in terms of ease of implementation. Finally, a prioritization list of actions are produced, which can be used to tackle the greatest issues in the shortest amount of time, optimized for impact.
Key takeaways:
Previous surveys on AAM public perception vary significantly in their results, from an acceptance rate of 45% to 83%. However, the average converges around 50%, indicating significant work necessary from all stakeholders to engage the public in a responsible, sustainable way.
Operations related to health & emergency services have the highest level of acceptance. On the other end of the spectrum are business & VIP ops, least accepted by the public.
The top 3 mitigation measures easiest to implement with the most impact are: 1) limiting minimum cruise altitudes, 2) establishing no-fly zones and 3) strategically placing vertiports.
The BFD: It’s notoriously difficult to quantify the subjective feelings of the public for a technology that does not widely exist yet. However, running surveys and workshops while talking to potential future users can quantify the opinions at a high level. This can enable regulators, infrastructure providers, vehicle designers and mobility operators to tailor their approach, aiming for favorable public perception from day 1. Social acceptance of AAM is key for operational scale and is on par with getting the tech right and creating a conducive regulatory environment.