Wu 1

Kaylee Wu

Professor Suzanne Richard McCarthy

ENGW 3315

29 July 2023

Software for Unmanned Aerial Vehicles in Disaster Relief Search for Survivors

1. Background

Unmanned aerial vehicles (UAVs), or more commonly known as drones, are an emerging

technology used by first responders to provide a birds-eye view of the aftermath of natural

disasters, such as earthquakes and floods, and conduct brief searches of inaccessible areas

(Papaioannou et al. 1). Recent research demonstrates that UAV technology has much more

potential in disaster relief. This paper will discuss current approaches to develop software for

UAVs to assist in the search for survivors after natural disasters, the limitations and concerns of

those approaches, and possible future research.

2. Current Research

The process of searching for survivors after all natural disasters involves assessing the

disaster area and locating survivors within the area, both of which can be optimized through

software algorithms. UAVs can and have been equipped with a variety of sensors to detect

survivors depending on the type of natural disaster and the accessibility of the sensor. Some of

the options include infrared and thermal cameras, temperature and carbon dioxide sensors, and

microphones (Alali and Alabady 3). Software techniques such as computer vision and artificial

intelligence have been used to develop algorithms that can detect survivors from the images and

data collected with sensors (Alali and Alabady 4-6). While sensors are useful, they are easily
 Wu 2

influenced by environmental factors, require close search to detect survivors, and do not address

the need to scan and search large areas without the knowledge of where to look (Oh and Han 1).

The next significant advance in UAV search technology seems to be automation, which can assist

in guiding UAVs to survivors in a more focused, safe search of disaster areas, as explored by

Sandino et al., Papaioannou et al., and Oh and Han. Automating the search process can provide

great support to the use of sensors in UAVs, as the UAVs will have an estimate of where to look

and use their sensors to confirm the locations of survivors.

As Sandino et al. recognizes, “Real-world emergency environments are dynamic,

complex, unknown, or partially unknown” (2). Therefore, to account for this, Sandino et al.

expands upon previous research and proposes adding cognition capabilities to UAVs, so the

drones are able to navigate through uncertain and unknown environments and make proper

decisions if they detect survivors, without the dependency on human control (3). The framework

creates UAVs with a set of modules with each module responsible for a different operation.

Vision-based sensors collect images of the environment (Sandino et al. 5). The computer vision

module collects and processes that data using artificial neural networks to determine whether a

survivor is detected in the frame, and if detected, the estimated location of the survivor and the

confidence of the detection (Sandino et al. 6). The decision-making module then uses a Partially

Observable Markov Decision Processes algorithm to develop the optimal sequence of actions to

accomplish the flight mission, which in this case would be to maneuver in a manner that makes

the view of the survivor more clear. Finally, the motion module controls the hardware to have the

UAV move accordingly (Sandino et al. 5). Many types of external sensors are used by the UAV

to estimate its position. This process allows the UAV to not only detect survivors, but also can be

applied to efficiently avoid obstacles and plan paths to locate survivors based on the certainty
 Wu 3

level of detecting them (Sandino et al. 9). Manual control of UAVs is inefficient and ineffective,

as there is no knowledge of where the survivor is, and even if the sensors detect a survivor, it is

difficult to pinpoint the exact location of where the survivor is. Additionally, maneuvering

manually through unknown environments is difficult. Thus, this proposed framework utilizes

software to enhance UAV navigation and survivor search.

Papaioannou et al. have a slightly different approach to automation for similar reasons.

Low-level mission constraints, mission objectives, and search mission specifications are

provided as input by a human operator to an algorithm that creates a search plan for the UAV to

follow. The mission constraints refer to the abilities and hardware of the UAV, including the

geometry of the sensor’s field of view. The mission objectives mainly refer to the mission

execution time and the energy efficiency of the UAV. The mission specifications refer to the

objects to be searched, the objects to be avoided as obstacles, and the probability of detecting

survivors (Papaioannou et al. 2). When deployed, the UAV will follow a path that optimizes

these inputs, search objects with a higher probability of detecting survivors more closely, quickly

scan objects that have a lower probability of detecting survivors, and avoid objects that resemble

obstacles (Papaioannou et al. 4). The purpose of this framework, as Papaioannou et al. states, is

“to augment the traditional SAR [search and rescue] missions with precise and efficient

automated search capabilities and with improved organization and planning”, as manually

operating a UAV in an optimal manner is “an error-prone process and requires a high degree of

human expertise” (17). This method is similar to the previous in that it allows the UAV to

navigate autonomously to avoid obstacles and search for survivors but differs in that it doesn’t

utilize artificial intelligence or require the UAV to make decisions. The UAVs merely follow the

programmed instructions.
 Wu 4

Oh and Han similarly recommend an automation approach to UAV search flights but

have a varying perspective regarding the sensors and how to detect survivors. Vision-based

sensors like cameras require close inspection of land, taking a long time to cover large areas, so

Oh and Han choose to utilize sensors that detect radio waves from mobile communication, WiFi,

and portable radio frequency units (2). The drones will launch in random directions, and when a

radio signal from a survivor is detected, the UAV will exchange the information with other UAVs

and the Ground Control System (GCS), which will all deploy their Smart Search module. The

module will estimate the survivor’s location with the Smart Search algorithm, and the UAVs will

generate and autonomously fly a direct path to the location while avoiding obstacles seen by the

flight algorithm (Oh and Han 6). As the drone gets closer to the estimated locations, it will

update that location based on Received Signal Strength Indicator (RSSI) and Time of Arrival

(ToA) data accumulated from the survivor, resulting in a more accurate location estimate (Oh

and Han 2). The GCS with a human control can also track the flight environment of the UAVs to

improve the accuracy and speed of the tracking, but this is not required, as the UAVs can

maneuver themselves (Oh and Han 6). The goal of this framework is “to expedite the search

operation in the disaster areas outside the mobile communication network where distressed

persons cannot request rescue”, as it functions in non-cellular and GPS shadow areas, which

improves upon previous research that cannot (Oh and Han 2). This approach is similar to the first

in that the UAV takes the lead in estimating the location of survivors and similar to the second in

that the human operator can influence the flight path by providing input.
 Wu 5

3. Limitations & Future Research

There are a few limitations and concerns with UAV automation that are shared across all

the mentioned approaches. The first and most obvious is the limit of the hardware. UAVs are

small, so they have a limit on how many sensors they can have and the size of the

microcomputers that are used to compute the search and flight algorithms (Alali and Alabady

14). UAVs are also battery powered, so there is a limit on flight time and the number of complex

tasks it can perform (Oh and Han 1). Current and future research must discover ways to optimize

what can be done within the bounds of the hardware.

Another concern that needs to be addressed is the integration of unmanned aircraft

systems (UAS) into the existing air traffic management (ATM) system to reduce the risk of

collisions with other aerial vehicles (Lieb et al. 1). Current regulations force UASs “to evade

manned aircraft or immediately initiate an emergency landing in order to guarantee their safety”,

hindering and interrupting UAS search and rescue missions (Lieb et al. 4). Therefore, to increase

air traffic safety and allow the UAS to perform their necessary tasks in disaster relief, it is

imperative to develop a system that takes into account the UASs and allows them to coexist with

other aerial vehicles in the national airspace. There is potential for future research and

advancement in this area to improve the communication and organization among aerial vehicles,

manned and unmanned, and develop an integrated crisis traffic management system that

minimizes collisions and allows all aerial vehicles to accomplish their missions (Lieb et al. 7).

Automating the communication and traffic management is an option to consider.

Ethics is also a concern in the automation of UAV search and flight operations. The use

of software and AI in disaster relief must be fair, transparent, and accountable (Gevaert et al. 1).

Without an effort to work towards these, AI cannot be considered ethical. Regarding fairness,
 Wu 6

“bias can also be the result of algorithm design or decision around metrics used to evaluate a

particular phenomenon”, as Gevaert et al. explains (2-3). With the mentioned approaches, the

data that is used to train the algorithms may be susceptible to bias, leading the algorithms to

make biased decisions. For example, in Oh and Han’s approach, the UAVs detect survivors based

on radio waves, but some vulnerable populations may not have devices that emit radio waves,

making it more difficult to detect them. Regarding transparency and accountability, it is essential

that those utilizing the AI and automation technology are able to understand, communicate, and

evaluate how the systems work, because “the introduction of AI techniques and their associated

complexity into disaster risk modeling processes may reduce the ability of government, the

public, and other important stakeholders to meaningfully participate in DRM [disaster risk

management]” (Gevaert et al. 3). If other involved organizations and stakeholders are unable to

understand how the UAVs are being used, the entire operation is at the mercy of those who

control and deploy the UAVs, or the UAV technology itself if no one is able to understand it. As

AI is able to make decisions and detect survivors, search and rescue teams may feel less

accountable and act with less intensity, because their decisions aren’t entirely their own so they

don’t have as much responsibility. Moreover, with AI being a new and promising technology,

there are high expectations for its potential, which may lead some to deploy and use immature

and untested technology (Gevaert et al. 3). Ethical use of AI and software in disaster relief is

important, and the stated considerations need to be addressed.

4. Conclusion

This paper discusses current research regarding software utilized by UAVs to search for

survivors of natural disasters. Three main studies are evaluated to demonstrate the variety of
 Wu 7

software solutions for UAV search tasks. These proposed search solutions, written by Sandino et

al., Papaioannou et al., and Oh and Han, collectively suggest introducing automation into the

search for survivors performed by UAVs. Following that, some limitations and concerns

regarding AI and automation are presented along with directions for future research to address

them. Overall, UAVs still have much potential to improve disaster relief.
 Wu 8

Works Cited

AlAli, Zahraa Tarik, and Salah Abdulghani Alabady. “A Survey of Disaster Management and

SAR Operations Using Sensors and Supporting Techniques.” International Journal of

Disaster Risk Reduction, vol. 82, 2022, https://doi.org/10.1016/j.ijdrr.2022.103295.

This research paper focuses on the role of sensors in disaster management. The

discussion includes research regarding the types of sensors used in detection of disasters and

survivors, the impact of sensors on the disaster management process, the supportive and assistive

use of sensors for other technologies, and advantages and disadvantages of using sensors in

disaster management. Authors Alali and Alabady have an educational and professional

background in mechatronics engineering and computer engineering, so this peer reviewed paper

has accuracy and validity. Furthermore, it specifies information regarding what sensors are

traditionally used with unmanned aerial vehicles, for what purposes, for which disasters, and

with which artificial intelligence techniques.

Gevaert, Caroline M., et al. “Fairness and Accountability of AI in Disaster Risk Management:

Opportunities and Challenges.” Patterns, vol. 2, no. 11, 12 Nov. 2021,

https://doi.org/10.1016/j.patter.2021.100363.

This publication analyzes potential concerns of using artificial intelligence (AI) in

disaster risk management (DRM) from an ethical and social perspective and considerations to

ensure AI is used responsibly in assisting societies in preparing for and recovering from natural

disasters. The main concepts examined in the text are fairness, accountability, and transparency.

Fairness refers to the mitigation of bias in the data and algorithms used in DRM. Transparency

and accountability refer to the necessity of understanding and communicating how the AI models
 Wu 9

work and how to evaluate the outputs of those models. The text suggests following these

guidelines to utilize AI ethically and inclusively in DRM but recognizes that much discussion

and research still needs to be done to fully address the issue. The authors of this source are

highly educated university professors from a variety of backgrounds, such as computer science,

earth observation science, philosophy, and geo-information, providing this source with

credibility and validity. Little published research has been done specifically regarding ethics in

AI for DRM, yet this is an important consideration when using AI, thus making this source

useful.

Lieb, Joonas, et al. “Identifying Needs and Requirements for an Integrated Crisis Traffic

Management (ICTM) Concept for Unmanned Aircraft Systems Supporting First

Responses in Crisis Situations.” 2021 IEEE/AIAA 40th Digital Avionics Systems

Conference (DASC), 2021, https://doi.org/10.1109/dasc52595.2021.9594475.

This study outlines the challenges and limitations of integrating unmanned aircraft

systems (UAS) into the existing national airspace system. Research done through workshops and

online meetings with Public Protection Disaster Relief organizations reveals other air traffic in

disaster areas threaten UAS missions operated by first responders, and communication and

coordination with other first responders is a challenge. As a result, the requirements and

specifications for a framework to integrate first responder UAS operations into the existing

airspace system, called integrated Crisis Traffic Management (iCTM), is developed in the text.

The new proposed system specifies air traffic rules and regulations for UAS first responder

missions, discusses required technologies and operational procedures to enable the framework,

and emphasizes the use of automating certain disaster relief UAS processes, such as the

supervision of UAS safety conditions. As this study was done recently by experts from the
 Wu 10

Institute of Flight Guidance or the Bavarian Red Cross, the information is relevant and accurate.

It provides a useful perspective on the current limitations of unmanned aerial vehicles in disaster

relief and conveys the potential and necessity of future research in integrating UAS into current

airspace systems.

Oh, Donggeun, and Junghee Han. “Smart Search System of Autonomous Flight UAVs for

Disaster Rescue.” Sensors, vol. 21, no. 20, 13 Oct. 2021,

https://doi.org/10.3390/s21206810.

This research article proposes an autonomous unmanned aerial vehicle (UAV) smart

search system to search for and locate survivors in disaster areas. In this system, the UAV detects

radio waves from WiFi, portable radio frequency terminals, and mobile devices and exchanges

the information to other UAVs and the ground controls server. The UAV then estimates the

survivor’s location and flies toward it without the need for human intervention or control,

collecting data from the signal to update its estimation of the survivor’s location. This process

repeats until the survivor is found. This proposed system speeds up the process of searching for

survivors and lowers the cost, as the UAVs do not need mobile communication networks, GPS,

or expensive cameras to pinpoint locations of survivors. Authors Oh and Han are educated in

electronics and information engineering and have performed similar research before, so the

source is valid. The text is useful in that it introduces a new method of locating survivors in

disaster areas which addresses limitations of previous research.

Papaioannou, Savvas, et al. “Towards Automated 3D Search Planning for Emergency Response

Missions.” Journal of Intelligent & Robotic Systems, vol. 103, no. 1, 3 Aug. 2021,

https://doi.org/10.1007/s10846-021-01449-4.
 Wu 11

This research paper aims to develop a framework that automates the planning and

execution of unmanned aerial vehicle (UAV) search missions in 3D environments. Based on

mission constraints, mission specifications, and mission objectives, the framework creates a

search plan that guides a UAV to search all the faces of a specific object of interest and avoid

collisions with obstacles without the need for manual control. The human operator would

provide all the details regarding a search mission to the UAV, but ultimately the UAV would be

able to navigate and identify objects, namely survivors, on its own. This proposal is unique and

builds upon previous research in that it allows the human operator to provide mission specific

information to the UAV to better inform the search for survivors, whereas previous research only

aims to avoid obstacles. Written by several highly educated experts in electronics engineering,

computer engineering, and telecommunications engineering and peer reviewed, this paper is

current and accurate. It offers a unique method for UAVs to perform search missions with more

precision and efficiency and introduces a concept that improves the organization and planning of

search missions compared to previous similar research.

Sandino, Juan, et al. “UAV Framework for Autonomous Onboard Navigation and People/Object

Detection in Cluttered Indoor Environments.” Remote Sensing, vol. 12, no. 20, 16 Oct.

2020, p. 3386, https://doi.org/10.3390/rs12203386.

This research article analyzes and tests the feasibility and utility of adding cognitive

capabilities to unmanned aerial vehicles (UAVs) in terms of natural disaster relief. With

cognitive capabilities, UAVs would not need supervision and the physical presence of a human to

function, so in the cases of uncertain, unpredictable environments, partial observability, and

victim detection, UAVs would have the ability to navigate autonomously. The new system

architecture extends previous contributions by running computer vision and decision-making

tasks onboard the UAV to avoid the dependency on communication modules to transfer and

process data to an external drive. An experiment in a controlled environment reveals the

potential of the system to increase the speed of monitoring damage and locating survivors, but

also demonstrates limitations that are necessary to address, including dynamic obstacles, an

increased dependency on the local position estimation system that makes the vehicle more prone

to collisions, and the endurance of the UAV. The authors, all specialized in UAV research, have

different disciplinary educational backgrounds, from computer science and technology to

electrical engineering, and provide innovative software developments to disaster relief

technology along with future suggested improvements, rendering the recent article valid,

accurate, and useful.