Unmanned Aerial Vehicles (UAVs), commonly known as drones, face many challenges and applications that can be improved through Artificial Intelligence (AI), edge computing, and edge AI.
AI “describes the workflow of machines that would require intelligence if performed by humans.” Recently, robots/drones have used AI to accomplish various intelligent tasks. Compared to traditional cloud computing, edge computing brings computing services (such as storage and processing) closer to the end users at the network edge. Computation is not done on remote cloud servers, but can be performed on nearby edge servers or devices (e.g., drones/unmanned aerial vehicles), or a combination of both. Edge AI can be summarized as the fusion of AI and edge computing.
The key technical challenges for drones include autonomous navigation, power management, security and privacy, formation control, computer vision, and communication.
Autonomous navigation can be further divided into localization and mapping, path planning, and collision avoidance systems. Important applications of drones include delivery systems, precision agriculture, civil infrastructure inspection, search and rescue operations, serving as aerial wireless base stations, and drone light shows.
Traditional cloud AI computing occurs on cloud servers, which offer more processing power and storage space compared to edge AI. However, edge AI has many advantages over cloud AI. Notably, it has lower latency, higher reliability, improved security and privacy, lower costs, and reduced energy consumption.
The identified technical challenges for drones have certain particularly important advantages of edge AI. For example, autonomous navigation (especially collision avoidance systems) largely relies on the lower latency of edge AI. Given the speed of drone flight, if obstacles cannot be detected and avoided within a fraction of a second, collisions are likely to occur. The improved latency provided by edge AI compared to cloud AI could be the difference between collision and no collision. Different applications also rely more on specific technical challenges. For instance, precision agriculture (such as crop management) mainly depends on the technical challenges of computer vision.
While the benefits of edge AI for drones and their technical challenges/applications are indisputable, there are still many challenges in implementing edge AI for drones. The key challenges of implementing edge AI for drones include developing distributed training algorithms, security and privacy, resource allocation, and real-time requirements. However, potential solutions can be identified for each implementation challenge. For example, the implementation challenge of “real-time requirements” outlines that drones often need feedback to be “real-time” (e.g., when tracking fast-moving cars).
However, sometimes collecting data, training models, and deciding actions can take too long. One solution is to reduce inference time through “model pruning” and “knowledge distillation.” Future research directions for edge AI applications in drones can also be defined by the challenges faced and lessons learned.
Drones and edge AI are exciting and active research fields.
The key technical challenges of drone systems can be divided into six categories: 1) Autonomous Navigation; 2) Formation Control; 3) Power Management; 4) Security and Privacy; 5) Computer Vision; 6) Communication.
A. Autonomous Navigation
1) Introduction:
Autonomous navigation can refer to vehicle navigation that is remotely operated by humans but has some simple onboard algorithms to take over and prevent collisions, or it can refer to fully automated vehicle navigation that reaches point A to point B without any human intervention (for example, drones used for package delivery). Depending on the application, vehicles autonomously navigate by adopting localization and mapping, path planning, and/or collision avoidance. For example, in the case of remotely operated drones using only a few simple onboard algorithms, only collision avoidance is adopted. In fully autonomous cases, localization and mapping, path planning, and collision avoidance are required. Therefore, this article introduces autonomous navigation from the perspectives of “localization and mapping,” “path planning,” and “collision avoidance systems.”
a) Localization and Mapping:
For robots, mapping is the process of constructing a map (2D or 3D) of a specific area, while localization is the process of determining the position and orientation of the robot relative to a reference frame. Accurate localization is a daunting task for any robot, and for aerial robots like drones, localization is especially difficult due to the three-dimensional nature of the environment. To make this accurate localization task easier, drones often heavily rely on Global Positioning System (GPS), so that GPS position measurements can be fused with onboard Inertial Measurement Unit (IMU) measurements to accurately estimate the drone’s posture (position and orientation). This method works well because GPS data can compensate for the cumulative errors of the IMU (caused by drift in its measurements).
However, in many cases, GPS services are unavailable/unreliable, such as indoors (in factories, warehouses, etc.), during emergencies/disaster situations, near tall trees/buildings, or near bodies of water. These environments are aptly referred to as GPS-denied environments, and navigating drones in such environments and achieving precise localization using only IMU measurements is very challenging due to the aforementioned cumulative errors of the IMU.
The most common solution to this problem is vision-based solutions, which accurately localize UAVs without GPS by combining IMU and visual sensor measurements. When the two measurements (IMU and visual sensor measurements) are fused, improved posture estimation can be achieved. The most famous of these vision-based solutions are Visual SLAM (Simultaneous Localization and Mapping) and Visual Odometry (VO). SLAM algorithms aim to estimate the robot’s posture while building a representation of the explored area (visual SLAM algorithms rely on visual sensors). VO predicts the UAV’s posture incrementally by analyzing the changes caused by motion on a series of images.
b) Path Planning:
Drone path planning is the problem of determining the path for a drone from a starting point to a target point. There are many types of path planning techniques, but they all attempt to find the optimal (or near-optimal) path based on performance metrics such as shortest time, shortest route, or lowest operational cost.
As expected, the integration of path planning and collision avoidance is crucial. They are often used together, so collision avoidance is often referred to as “local path planning,” while path planning is referred to as “global path planning.” The idea is that global path planning generates the best route while considering the entire environment, while local path planning performs local processing when environmental changes are detected and executes collision avoidance actions accordingly. Once collision avoidance actions are executed, an attempt is made to return to the global path.
Drone path planning (global path planning) techniques can be divided into various types. One example is the research by Aggarwal and Kumar, which categorizes path planning techniques into representative techniques, cooperative techniques, and non-cooperative techniques.
c) Collision Avoidance Systems:
As drones fly in the air and are often very fast, they face a high risk of collisions, whether they collide with other objects or other objects collide with them. Collision avoidance systems are crucial for drones to avoid such collisions and ensure safe flight.
Yasin et al. classify collision avoidance systems into two main categories: 1) Perception and 2) Action, with perception being primary. In the perception phase, sensors perceive the environment and detect obstacles, while in the action phase, collision avoidance methods utilize this information to avoid collisions. Perception is subdivided into active sensors (sensors that emit/launch and then receive signals from their own source) and passive sensors (sensors that only read energy released by other sources). Active sensors are further divided into sonar, LiDAR, and radar, while passive sensors are divided into cameras and infrared.
In the action phase, collision avoidance methods are divided into “four major methods”: 1) Geometric; 2) Optimization; 3) Force Field; 4) Perception and Avoidance. The “Geometric” collision avoidance method uses the speed and position information of the drone and obstacles to ensure that minimum distance violations do not occur, often involving trajectory simulation. The “Optimization” collision avoidance method uses known obstacle parameters to find the best (or near-optimal) route. The “Force Field” collision avoidance method manipulates attractive/repulsive forces, while the “Perception and Avoidance” method makes avoidance decisions based on obstacle detection in real-time.
2) The Role of AI:
The greatest contribution of AI to autonomous navigation is the computer vision algorithms, which are crucial for the three major challenges: 1) Localization and Mapping; 2) Path Planning; 3) Collision Avoidance Systems. Recent advances in computer vision algorithms (especially deep learning algorithms) and camera technologies (especially RGB cameras) enable drones to effectively detect objects using high-resolution, lightweight, and inexpensive onboard cameras. In localization and mapping, especially in the absence of GPS data, there is a heavy reliance on the use of computer vision algorithms in visual methods (e.g., SLAM and VO) as discussed in Section III-A1a. Additionally, in path planning and collision avoidance systems, computer vision algorithms are crucial for quickly detecting oncoming objects to avoid collisions.
3) The Role of Edge AI:
Using edge AI can bring many benefits to autonomous navigation compared to traditional AI under cloud models. While all general advantages of edge computing (lower latency, higher reliability, improved security and privacy, reduced costs, and reduced energy consumption) and all general additional advantages of edge AI (further improvements in energy consumption/privacy/communication) apply here, the lower latency and higher reliability offered by edge AI compared to traditional cloud AI are particularly important for autonomous navigation problems. Therefore, this section briefly discusses the advantages and disadvantages of edge AI.
a) Lower Latency:
Using AI under traditional cloud computing may involve significant transmission latency in sending and receiving data between the drone and the central cloud server. Edge AI allows most processing to occur on terminal devices or nearby edge servers, resulting in much shorter associated transmission latencies compared to remote clouds. Even in the situations described in Section II-B, where some applications require edge servers to communicate with core clouds for data synchronization, the transmission latency is not significant because preprocessing at the edge reduces overall traffic load (i.e., less data sent from each device/drone).
Collision avoidance systems are a typical example of real-time latency-sensitive challenges that particularly rely on low latency, making edge AI potentially essential. The process of detecting and avoiding oncoming objects (unless the drone and the oncoming objects are moving very slowly) needs to be completed within a fraction of a second. If this process takes too long, it could result in crashes, leading to significant economic losses or worse (e.g., fatal consequences). For path planning, localization, and mapping, the latency of traditional AI may be sufficient in most static environments, especially in sparsely populated or high-altitude areas. Nevertheless, considering the limitations of high latency on flight speed, the improved latency provided by edge AI remains a significant benefit.
b) Improved Reliability:
Running AI under traditional cloud computing relies on the reliability of central cloud servers. If the central cloud server is inaccessible/fails, it can result in catastrophic consequences (e.g., crashes). The likelihood of this happening with edge AI is much lower because processing occurs closer to the user (on edge servers/devices), which significantly reduces the chances of network interruptions. If an edge server does fail, another server should be able to provide service; even if not, the terminal devices can handle a considerable amount of requests on their own and still access some locally stored data.
When using AI under traditional cloud computing models, the unavailability/outage of the central cloud server can lead to crashes because the drone’s video source cannot be analyzed, and oncoming objects that could collide with the drone cannot be detected. Path planning, localization, and mapping can cope with slightly longer times without the central cloud server using traditional cloud AI, but if this loss of connection persists for a long time, hovering in the air without completing tasks or crashing remains inevitable.
c) Disadvantages:
Edge AI has two significant disadvantages that affect autonomous navigation: there are significantly fewer network devices in certain places, and there are fewer technicians available to implement, repair, or manage network devices in those areas. In other words, in sparsely populated areas and/or areas with limited financial or technical resources, there may be fewer edge servers on the network, and in many such areas, there may be fewer technicians capable of implementing, repairing, and managing edge networks. This can affect autonomous navigation, as it may rely on edge servers, meaning the positive impacts of edge AI may diminish, especially in terms of latency and reliability.
Moreover, another disadvantage of autonomous navigation is the need to offload drone tasks between different edge servers. When a drone leaves the service range of a specific edge server, that server may need to migrate the offloaded tasks of the drone to other edge servers. Due to the high speed of drone flight, this can limit the task performance of the drone.
B. Formation Control
1) Introduction:
When performing drone tasks, multiple drones working together is often preferable to a single drone completing tasks alone (see the end of Section II-A). Driven by the advantages of multiple drones working together (such as transporting large payloads or searching for objects/people over large areas), extensive research has been conducted in recent decades on formation-related studies, of which formation control is the most active research topic.
Formation control is inspired by self-organization in nature, such as flocks of birds, and is the coordinated control of multiple robots in “formation.” Formation is defined as a network of interconnected agents regulated by controllers, where each agent must maintain relationships with neighboring agents. Formation control can be roughly divided into two categories: 1) Leader-Follower and 2) Unleadered.
2) The Role of AI:
a) Reinforcement Learning:
Reinforcement Learning (RL) is a field of machine learning and thus also a field of AI. It can be described as learning how to maximize a numerical reward signal, where the learner is not given actions to take but must try these actions and infer which actions yield the greatest reward. RL is particularly useful for the coordinated movement and communication of drones. Therefore, many papers, especially recent ones, have studied the application of RL in drone formation control. For example, Knopp et al. proposed an RL algorithm using GQ (λ ) for leader-follower formation control scenarios.
b) Deep Reinforcement Learning:
One problem faced by RL is that it can be very challenging for RL algorithms to learn from each state and determine the best reward path (i.e., strategy). RL can be combined with deep learning to produce another subfield of machine learning called Deep RL (DRL). DRL can use neural networks to handle higher-dimensional state/action spaces. For example, Conde et al. used DRL to drive multiple drones to achieve formation, allowing the use of deep neural networks to estimate the quality of specific states.
c) Multi-Agent Systems and Reinforcement Learning:
Multi-Agent Systems (MAS) are a method for solving complex problems by breaking them down into smaller tasks. Each individual task is assigned to an autonomous entity called an “agent,” and each agent selects appropriate actions to solve the task based on multiple inputs. Inputs include their operational history, interactions with neighboring agents, and their goals. RL can be effectively applied to such MAS. For example, Liu et al. proposed a RL-based distributed model-free solution to solve the MAS leader-follower formation control problem.
3) The Role of Edge AI:
Similar to the case of autonomous navigation, formation control has many benefits when adopting edge AI instead of traditional cloud computing models. Therefore, most general advantages of edge computing (lower latency, higher reliability, etc.) and general additional advantages of edge AI (e.g., further improvements in energy consumption) apply. Similarly, this section will focus on particularly interesting and important advantages, which for formation control problems is reducing costs.
a) Lower Costs:
When referring to drone formations, we mean swarms of drones that generate large amounts of streaming data that is continuously transmitted from the drones for processing. Edge AI can reduce the amount of data that needs to be sent to the server through onboard preprocessing and can also reduce the distance data is sent by sending it to edge servers (instead of more remote centralized cloud servers). The reduced data volume is particularly helpful in lowering operational costs, as cloud computing costs can be quite high when processing large amounts of data. That said, the costs of deploying and maintaining edge devices spread over vast areas can be quite high and should be considered.
C. Power Management
1) Introduction:
A well-known issue with drones is their limited battery life, leading to limited flight times. In addition to improving the energy efficiency of drones, more energy can be obtained through innovative methods (i.e., energy harvesting technologies) to extend battery life. Common energy harvesting technologies include solar energy, wireless charging, and battery replacement.
a) Solar Energy:
Solar energy is cost-effective and environmentally friendly, making it an ideal energy source. Solar panels are particularly useful for drones, as they need to fly for extended periods at high altitudes. Solar-powered drones typically use fixed batteries as auxiliary power, usually not overly relying on batteries during the day, but battery power becomes crucial at night or in adverse weather conditions. That said, surface area, weight limitations, and dependence on light intensity are major limiting factors for commercial solar-powered drone applications.
b) Wireless Charging:
Wireless charging has not yet become a standard feature for drones. However, in recent years, significant research has been conducted to make wireless charging for drones possible. Explored technologies include capacitive coupling-based technologies, magnetic resonance-based technologies, and even charging from power lines through wireless power transmission.
2) The Role of AI:
AI can address power management issues through planning and optimization algorithms. For example, algorithms that find the best route for drones to minimize power consumption or select the best charging/battery swapping stations, taking into account current battery levels and deviations from the current route. For instance, Zhang et al. used DRL to find the fastest path to a charging station.
Additionally, Convolutional Neural Networks (CNNs) can be used to identify charging stations and their landing points, while Recurrent Neural Networks can accurately predict when a drone’s battery will run out, and AI can also help improve the efficiency of drones, allowing them to achieve more goals with less power.
3) The Role of Edge AI:
As with Sections III-A3 and III-B3, the general advantages of edge computing and general additional advantages of edge AI also apply here. The advantage of edge AI in reducing energy consumption is particularly important and will be discussed.
a) Reducing Energy Consumption:
As described in Section III-B3, edge AI can reduce the amount of data that needs to be sent to the server through onboard preprocessing and can also reduce the distance data is sent by sending it to edge servers (instead of more remote centralized cloud servers). In addition to lowering costs, this can also reduce energy consumption, as transmitting data to remote clouds incurs significant energy costs. By reducing the size of the data and the distance of transmission, overall energy consumption can be lowered, which is especially important for drones as batteries are a scarce resource. Furthermore, when using edge AI, distributed federated learning within drone swarms can further reduce the data sent to servers and the associated energy consumption. For example, Zeng et al. proposed a drone swarm framework based on distributed federated learning with a “leader drone” and several “follower drones.” All follower drones locally train models based on their collected data, and then each follower drone sends this locally trained model to the leader drone. This drone aggregates all models to create a global model and then sends it to the subsequent drones.
D. Security and Privacy
1) Introduction:
Security and privacy are very important issues for all digital systems, especially for drones. Compared to other invasive devices, drones have unique flexible access technologies (which are attractive to criminals) and are also vulnerable to attacks targeting wireless links, network elements, physical elements, and the interfaces between network and physical elements.
a) Security:
While many papers discussing drone security address the same issues, the approaches taken differ.
Shakhatreh et al. classify the attack vectors of drone systems into attacks on communication links, the drones themselves, ground control stations (GCS), and humans. They also define the three major cybersecurity challenges for drone systems: 1) Confidentiality challenges (protecting information from unauthorized access); 2) Availability challenges (ensuring that drone systems and data work as intended and are accessible to authenticated users); 3) Integrity challenges (ensuring the authenticity of information). Ullah et al. briefly explain the classification of attackers (internal/external, malicious/rational, active/passive, and local/extended) and attacks (e.g., spoofing, denial of service, and GPS spoofing). However, this section analyzes drone attacks from the perspective of sensors and communication links.
The information acquired by sensors can change the behavior of drones and have significant implications for security. GPS sensors are a common attack sensor because they are often used to obtain accurate location information. The most common ways to attack drone GPS are “jamming” or “spoofing.” Jamming refers to unauthorized parties broadcasting interference signals to prevent the reception of various signals. Spoofing can involve unauthorized parties recording satellite signals and transmitting them to the drone (relay-type spoofing) or using certain programs to generate signals based on real signals (generation-type spoofing). One method to address GPS jamming is to adopt alternative navigation methods, such as using visual and inertial navigation systems that employ SLAM or VO (see Localization and Mapping section, Section III-A1a). Solutions to GPS spoofing include authenticating GPS signals, such as checking GPS observables that represent signal travel time or detecting sudden changes in signal power/observables that may indicate the onset of spoofing attacks. Other commonly vulnerable sensors include binocular vision sensors (which can be spoofed, e.g., by using lasers pointing at the ground to induce features) or MEMS gyroscopes that may be attacked by ultrasound (e.g., causing the drone to lose balance).
In terms of communication links, the interaction between drones and ground stations relies on communication links that are insecure and susceptible to attacks. For example, one of the ways drones communicate with ground terminals is via WiFi, which is susceptible to Deauth attacks that disrupt the connection between the drone and the terminal, compromising the drone’s control passwords. Solutions include having the drone ask the user whether to automatically return after losing contact for 10 seconds, or using radio signals instead of WiFi signals.
b) Privacy:
Drones can easily invade privacy, and it is challenging to capture invading drones. There are two main solutions to prevent drones from invading personal privacy. One is to register home addresses in a no-fly zone database, but this still does not ensure that invading drones will not fly into no-fly zones. The second solution is to use technologies/systems to detect, track, and deploy drones in space.
Additionally, malware can exploit drones to collect personal information. For example, Snoopy malware can be installed on drones to collect personal information and track/analyze individuals using smartphones. Drones need to continuously evolve to counter this ever-evolving malware.
2) The Role of AI:
AI can be used to help prevent drones from being attacked, as well as in the systems discussed above to prevent drones themselves from invading people’s privacy (by flying over/near no-fly zones). Regarding the latter, by 2026, the “global anti-drone market” is expected to reach $2.597 billion. A surveillance system by Japan Electric Company is an example that uses acoustic, thermal, infrared, and/or radio communication sensors/detectors to perceive invading drones and provides options for tracking system owners to discard/capture drones. Algorithms for detecting drones particularly employ AI, such as the drone detection algorithm based on Artificial Neural Networks (ANN) proposed by Zhang et al.
For the former (preventing drones themselves from being attacked), Challita et al. believe that it is important to detect potential attacks by discovering bad/abnormal drone movements. Challita et al. use RNN as an example, where RNN can predict the movements of drones, thus detecting abnormal movements of drones. Additionally, AI protection can utilize models trained against malicious files to protect drones from zero-day attacks.
3) The Role of Edge AI:
As with previous sections on the role of edge AI for other technical challenges, the general advantages of edge computing and the general additional advantages of edge AI also apply. Clearly, the security and privacy advantages of edge AI are partly relevant and will be discussed.
a) Improved Security:
Typically, using AI in traditional cloud computing models requires transmitting all data to a central server. This is considered “highly vulnerable” because a single attack (e.g., DDoS attack) can cause severe damage. Using edge AI means that data processing is distributed across multiple devices/servers. While it must be acknowledged that this distribution of data processing increases the potential attack surface, the risks are more dispersed, thus reducing the impact of successful attacks (such as DDoS attacks). Furthermore, since edge AI supports processing at the edge, less data is sent, so there is less data that can be intercepted.
In terms of practical shutdown attacks, the distributed and decentralized nature of edge computing means that the vulnerable parts of edge AI systems are easier to shut down than traditional cloud AI systems, which often require shutting down the entire network.
Integrating blockchain into drone systems can provide many security advantages. For example, blockchain can ensure that each drone has a copy of the blockchain, thereby reducing the vulnerability of the drone network to signal interference. In this case, individual drones use the details of other drones’ flight paths contained in their blockchain copies to determine their paths. Another example is how blockchain can detect malicious drones that change network information. Malicious drones may have initially been part of the drone network and later been hijacked or entered the network later. Blockchain can help prevent such attacks through blockchain consensus algorithms, where any drone can report suspicious activities. If the number of entries contradicting the drone exceeds a certain threshold, that drone is labeled as malicious.
b) Improved Privacy:
Edge AI enables real-time computation. If it is inevitable to capture/record images that invade privacy, data can be processed locally on the drone or on edge servers without sending and storing it on remote centralized cloud servers. The significance of this is that such data is less likely to be hacked and does not have to be stored on remote centralized cloud servers.
Federated learning and blockchain can both enhance the security of drone communications. Federated learning can avoid sending any raw data from devices/drones (only local model updates are sent). Blockchain can encrypt data and store it in the blockchain so that no one can access the data without the correct decryption key. Additionally, blockchain can protect four main types of data in the drone network (drone identifiers, flight path controls, sensor data, and flight schedules) by writing and updating them within blockchain blocks.
c) Disadvantages:
As mentioned in the security section (Section III-D3a), it should be noted that distributing data/data processing across multiple edge nodes has the disadvantage of increasing the potential attack surface. Even with blockchain-based systems, each drone has a copy of the distributed ledger, meaning that some sensitive information propagates to all drones in the system. Additionally, it should be noted that in a blockchain-based drone swarm, if more than half of the drones in the swarm are hacked, the entire swarm can be controlled (this attack is known as a 51% attack).
E. Computer Vision
1) Introduction:
The purpose of computer vision is to enable computers to understand the environment from visual information (whether from a single image or a series of images). In recent years, there has been growing interest in the automatic understanding of visual data collected by drones, as computer vision plays a crucial role in most drone applications (from aerial photography to SAR operations).
From the perspective of computer vision, the core task of such applications is scene parsing. Different applications require different levels of scene parsing, from locating objects to determining precise boundaries of objects to recognizing objects. Drone computer vision applications include object detection, object recognition, object tracking, collision avoidance, autonomous navigation, and 3-D reconstruction. Such image processing can be done remotely on servers (either edge or central cloud) or on the drone (embedded).
a) Remote Computer Vision Processing:
Drones typically lack the processing power to handle images captured by drone cameras, so processing needs to occur at different locations. From a latency perspective, the ideal processing location is an edge server, although computer vision can also be processed on more distant centralized servers.
b) Real-Time Embedded Computer Vision Processing:
If the goal is to make drones truly autonomous and reliable, real-time embedded computer vision processing is preferable to remote computer vision processing, as remote processing requires high bandwidth, minimal latency, and extremely reliable wireless links, which cannot always be guaranteed.
The most prominent limitation of real-time embedded computer vision processing for drones is onboard computing capability. Van Beeck et al. point out that the computational requirements of state-of-the-art drone computer vision algorithms often conflict with hardware resource limitations.
2) The Role of AI:
In computer vision technology, the use of AI is not mandatory; for example, Petricca et al. can perform corrosion detection based on the number of pixels containing certain red components. Nevertheless, despite the fact that AI-based techniques may require large datasets to achieve optimal results, computer vision applications heavily utilize AI. The area of AI that intersects most with computer vision is deep learning.
a) Deep Learning:
According to research by Lecun et al., deep learning enables computational models composed of multiple processing layers to learn data representations with multiple levels of abstraction. The early layers aim to learn low-level features such as edges, while the later layers aim to combine features into more complete representations. One example of the application of deep learning in drone computer vision is the method proposed by Ye et al., which employs a deep learning classifier to detect and track other drones. Another example is the method proposed by Padhy et al., which uses a CNN model to help drones navigate autonomously in GPS-denied indoor corridor environments.
3) The Role of Edge AI:
The role of edge AI in computer vision involves computer vision processing on edge servers and drones. When AI processing can be done on edge servers or devices (rather than under traditional cloud models), all general advantages of edge computing and additional advantages of edge AI apply. The advantage of low latency is particularly important for computer vision applications, so this will be emphasized. As with the role of AI section (Section III-E2), this section will also particularly highlight deep learning, focusing on embedded deep learning, which has not been discussed yet.
a) Lower Latency:
As mentioned in Section III-A.6, drones using traditional cloud AI experience considerable transmission latency when sending and receiving data between them and the central cloud server. Edge AI allows most processing to occur on terminal devices or nearby edge servers, resulting in significantly shorter transmission latencies compared to remote clouds. Even in situations where some applications require edge servers to communicate with core clouds for data synchronization, the overall traffic load is reduced due to preprocessing at the edge, so transmission latency is not significant.
Most computer vision applications (from object detection to object recognition to object tracking to collision avoidance to self-navigation) require low latency because if certain latency standards are not met, objects may be lost or collide.
b) Deep Learning:
According to Van Beeck, at the “UAVision2020” workshop focused on real-time image processing on drones, all accepted workshop papers (covering a wide range of different applications) utilized deep learning. Castellano et al. and Zhao et al. are good examples of using deep learning to describe crowd counting or understanding. Other examples include Stad and Zhang using deep learning for object tracking and Peralta using deep learning for 3-D reconstruction. Recent examples include Onishi and Ise using CNN methods to build a tree recognition and mapping system using drone RGB images, and Kung et al. proposing a CNN model for automatic detection of building defects (such as cracks).
Source: Military Eagle DynamicsEditor-in-Chief: Wu MinDeputyEditor-in-Chief:Yu WenjingOn-Duty Editor: Zhang ShuaiweiEditor: Sun Yiwen Proofreader: Liu Jianshu
