In this blog post, we will examine the basic structure and operating principles of unmanned vehicles and autonomous vehicles, and explore the possibility of multiple unmanned vehicles sharing information and driving cooperatively—safely and efficiently—by utilizing swarm intelligence inspired by the swarming behavior of animals.
The Evolution of Unmanned Vehicles and Autonomous Driving
Imagine a driver letting go of the steering wheel on a highway. On a national highway where many vehicles travel at speeds exceeding 100 km/h, such behavior is tantamount to suicide. So, does that person really want to die? Or is that car some kind of “Transformer” with intelligence and judgment? The answer lies in unmanned vehicles. It has long been predicted that in the future, vehicles will no longer be manually operated by humans but will instead be automated to move on their own. This technology is no longer merely a vision of the future. Autonomous driving technology is being tested on actual roads, and services featuring driverless operation have already emerged in restricted areas. However, current autonomous driving technology has not yet reached the stage where it can completely replace human driving on all roads and in all situations. The U.S. National Highway Traffic Safety Administration (NHTSA) classifies driving automation into Levels 0 through 5. Currently, Level 2 driver-assistance systems—which require the driver’s constant attention—are the most common in vehicles available to consumers. In contrast, Level 3 through 5 automated driving systems are still primarily focused on limited testing, development, and operation under specific conditions.
How do unmanned autonomous vehicles operate?
An UGV (Unmanned Ground Vehicle), or unmanned autonomous vehicle, refers, as the name implies, to a vehicle capable of assessing its surroundings, making decisions, and taking action without direct human control. Broadly speaking, this term can encompass various types of vehicles not directly operated by humans—such as those used for military operations, general purposes, or research—but here it is used specifically to refer to vehicles that can safely navigate general road environments alongside other vehicles without direct human driving.
An UGV consists of the mechanical components that make up a basic vehicle, along with sensors for obstacle detection, vehicle-to-vehicle communication devices, an onboard navigation system, and the technology that integrates these elements to collect information and control the vehicle. To draw an analogy with humans, these components correspond to the sensory organs that perceive the surrounding environment, the means of language and communication, and the brain that determines actions and judgments. Using sensors such as laser scanners, vision cameras, and GPS, it determines the speed and location of other vehicles nearby, as well as the terrain and road conditions; shares information regarding destinations and driving-related details with other vehicles via communication; and decides on and executes driving strategies based on the information received. Today’s autonomous driving systems also combine various sensors—such as cameras, radar, LiDAR, and satellite navigation—with computing technologies to perceive the surrounding environment and control the vehicle’s operation. However, while sensor performance is certainly important, what truly matters is which algorithms are used to determine the vehicle’s behavior under given conditions. This is because an autonomous driving system does not simply stop at detecting objects in its surroundings; it must also determine what actions to take next based on the information it has recognized.
The success or failure of UGV control depends on whether the vehicle can stay within designated lanes, avoid colliding with obstacles, and travel as safely and quickly as possible. Currently, on roads with parallel lane markings where vehicles are arranged in a line, the driving strategy can be explained relatively simply.
The concept boils down to “slow down when the distance to the vehicle ahead becomes too close, and accelerate when the path ahead is clear.” A significant number of algorithms related to this have already been developed. However, in environments where conditions become more complex—such as intersections or highways—in environments where lane markings are unclear, and in situations where multiple vehicles are distributed in a near-random manner, the problem becomes far more complex because each vehicle must drive while taking the movements of the others into account. In particular, cooperative driving—where multiple autonomous vehicles share information such as their positions, speeds, and directions of travel to move as a single group—presents challenges distinct from those of individual autonomous vehicles. Therefore, the advancement of cooperative driving technology is crucial for achieving a higher level of commercialization and stable operation of autonomous vehicles, and research and development in this area are ongoing. One source of inspiration for solving these problems is the social behavior of animal herds.
What are swarm behavior and swarm intelligence in animals?
Many animals do not live as isolated individuals but instead gather in groups for several reasons. The collective behavior and intelligence that emerge from this group living are referred to as swarm intelligence. Animals that live in swarms confuse predators and better identify their locations, thereby increasing their chances of survival. In particular, among birds, fish, and insects, it can be observed that while individual members appear to exchange information in complex ways, they actually follow relatively simple behavioral rules, resulting in orderly collective behavior as a whole.
There are dozens of species that demonstrate swarm intelligence. The most representative examples include birds, fish, and insects. Birds are well known for flying in a V-shaped formation. This formation arises because, when multiple birds fly together, the movements of the individual in front influence the flight of those following behind; since the birds can move efficiently by coordinating their movements, this is considered a prime example of swarm behavior.
Fish that live in schools in the ocean have also long been a subject of fascination for researchers. Many scholars have studied how fish are able to swim in the same direction so quickly and nimbly while maintaining a consistent distance from one another and avoiding collisions. One such study explains that fish within a school move according to the following three basic principles. First, they swim side by side in the same direction as the surrounding fish. Second, if an individual gets too close to another, they move away from each other. Third, they seek to approach fish that are outside a certain range. Simulation results based on these principles closely resembled the actual movements of fish schools, suggesting that seemingly complex swarm behavior can be explained by a combination of relatively simple behavioral rules. This model of swarm behavior has since evolved into a key concept used in various fields, such as computer graphics and robotics.
This is similar to love fostered through a healthy romantic relationship. For lovers, looking in the same direction, sharing similar thoughts, building memories together, and forming a bond of mutual understanding are essential elements of love. However, if one partner tries to share too much, it is sometimes necessary to gently push them away. This is because attempting to know and control everything turns into obsession rather than love, which can diminish the mutual attraction between the two. Conversely, it probably goes without saying that before affection fades or the couple grows distant, efforts to reconnect—such as being affectionate or giving gifts and planning special events—are necessary. By adhering to these three principles, you can naturally nurture a healthy and beautiful love.
The Principles of Cooperative Swimming Learned from Fish Schools
Recent studies have shown that even if each individual does not consider every surrounding individual one by one, but instead applies rules such as avoidance and approach only to certain individuals within a certain range, collectively aligned group behavior can still emerge. In particular, the Boids model—one of the representative models explaining swarm behavior—shows that individuals combine three behaviors—separation, alignment, and aggregation—based on their relationships with neighboring individuals.
Separation is the behavior of avoiding getting too close to neighboring entities; alignment is the behavior of moving in a direction similar to that of neighboring entities; and aggregation is the behavior of moving toward the average position of neighboring entities. The key point is that each entity does not control the entire swarm but rather reacts to its limited set of immediate neighbors.From this perspective, cooperative driving among unmanned vehicles can be conceptualized in a similar way. Rather than each car having complete information about every other car on the road and a central authority controlling the movement of every vehicle individually, each car detects the position, speed, and direction of movement of the vehicles in its immediate vicinity—with which it directly interacts—and adjusts its own movement accordingly. If the distance to the vehicle ahead becomes too close, it slows down; if an appropriate distance is maintained from surrounding vehicles, it moves in the same direction; and it adjusts its direction of travel based on its relationship with vehicles or groups that have moved too far away. When this individual behavior occurs simultaneously across multiple vehicles, the collective movement can appear orderly, resembling a single swarm.When explaining the swarming behavior of fish, we can distinguish between how surrounding objects are perceived and reflected in behavior, as seen in the Metric Model and the Topological Model. In the Metric Model, a fish’s behavior is based on its physical distance from all surrounding fish. In contrast, the topological model explains behavior as being determined based on a fixed number of the nearest objects, regardless of distance. This perspective is also significant for cooperative driving among unmanned vehicles. If the positions of all vehicles must be constantly monitored, the amount of information to process increases dramatically as the number of vehicles grows; however, if decisions are based solely on nearby vehicles that directly affect the individual vehicle, behavior can be determined more efficiently.Of course, cooperative driving by actual unmanned vehicles is far more complex than the movements of fish or birds. Cars travel at high speeds, are subject to clear constraints such as roads and traffic laws, and must simultaneously account for various factors, including pedestrians, obstacles, traffic lights, and intersections. Furthermore, since it cannot be guaranteed that vehicle-to-vehicle communication will always be flawless, technology is needed to appropriately combine information collected by sensors with information received from other vehicles and to select safe actions even in uncertain situations. Nevertheless, the principle by which simple rules and individual judgments—as seen in the swarm behavior of animals—create an overall orderly movement can serve as an important reference for researching cooperative driving involving multiple autonomous vehicles. Although autonomous driving technology has advanced to the point where actual operation and testing are taking place under limited conditions, fully autonomous driving is not yet possible in all environments, and challenges related to safety verification as well as technical and regulatory issues remain. Therefore, in future research on unmanned vehicles, it is important to study not only the autonomous decision-making of individual vehicles but also methods by which multiple vehicles can move safely by sharing information and cooperating with one another.
