Details of Research

Structure of R&D Themes Under the Project

Evolvable swarm robots with shared awareness
to settle unknown and explored areas

Evolvable network intelligence systems

We will develop systems for robots to share intelligence in a group and act strategically based on the environment. This system will enable robot swarms to engage in autonomous decision making, adapting flexibly and efficiently to unknown environments. Network intelligence is a key foundation for robots to exchange information with one another and take the optimal actions to suit the environment.

Engineering behavioral controls and networked intelligence in evolvable swarm robots

We will study ways for individual low-functionality robots to autonomously form groups without having recording environmental information and individual behavior, and without controlling the position of individuals.
Swarm sizes and behaviors that are appropriate for the spaces and tasks involved will be autonomously determined, improving spatial exploration abilities for unknown environments, and enhancing the ability to pass through confined spaces with changes in behavior patterns and jumping movement.
We will study guidance systems and ways to guide inter-swarm cooperative behavior for individual robots, single swarms and multiple swarms, by moving real markers that physically exist, as well as virtual markers created as an aggregate of multiple real markers.

Controlling evolvable network intelligence through the analysis of swarm-collected information

We will pursue the development of network intelligence capable of managing the measurement and activity data acquired by individual robots and swarms, and understanding the status of robots, environments and other factors by analyzing that data, thereby building analysis and database functionality.
When unknown areas are explored, the positional data on each individual robot and the swarm contains errors, making the information ambiguous.
To address this, we will design spatial information management methods that handle interconnection information between observation sites geometrically rather than using coordinate information, developing intelligence that can be utilized for recognizing anomalies and danger based on swarm behavior and other factors.

Design and implementation of autonomous distributed network intelligence with evolution and coevolution

To realize exploration, surveying, container transportation and other activities within a lunar lava tube for which no environmental information has been obtained in advance, AI technologies such as multi-agent reinforcement learning will be used, designing network intelligence that will autonomously determine the roles and behavior of individuals and swarms, along with the necessary functionality.
To ensure adaptability to challenging environments, these functions will be designed as algorithms and systems that are able to take into account power consumption, communication distances, behavior under asynchronous conditions and other factors, and allocate roles within a swarm to handle individual performance differences, changes over time, damage and other eventualities.

Realizing individual evolution and swarm coevolution functions

We will build mechanisms for robot swarms to continually evolve, while developing technologies for individual robots to learn from one another while enhancing the swarm’s capabilities as a group.
This will ensure that the robots have adaptive capabilities, enabling them to respond to a changing environment.
In the process of evolution, the robots’ functions will be changed flexibly, enabling them to take on new missions.
These technologies are the key to supporting long-term activities.

Improving the flexibility of control functions and speeding up processing for individual evolution and swarm coevolution

To enable the flexible expansion and update of the functions installed in the robots, a computing architecture implementing mechanisms to connect functional modules and implement highly granular tasking at the hardware level will be developed.
The functions installed in the robots will be made updatable through the addition or reconfiguration of modular connections, achieving individual evolution functions in the control equipment.
In addition, by using electromagnetic waves to build robust communications between individual robots and information networks, we aim to realize a system that can be implemented in individual robots along with an evolvable control unit.

Designing and implementing evolvable control units with low power consumption architecture and shared networks for high processing speeds

We will research the computing circuit architecture making up the control units of the small robots, considering the balance between processing speed and power consumption to enable advanced information processing for images, AI and so on.
We will also design methods to achieve data processing and administrative functions associated with task execution in a modular connection format as part of R&D into the control units.
Additionally, we will develop evolvable control units that enable the network sharing of the small robots’ functional modules.

Achieving individual evolution and swarm coevolution functions through data flow control for highly granular tasking

Inside individual robots and swarms, we will study and design data flow control, including the control of connection data between modules and the modification, addition or deletion of connections, on software and simulators. We will then integrate this with the hardware layer to develop evolvable control units.
In addition, we will implement support functions through AI technologies, human interface technologies and GUIs within the software environment to manage and operate the functions and shared data of individual robots, swarms and network intelligence, and study the application of network intelligence in the lunar information spaces that could appear in the future.

Realizing a modular sharing and connection mechanism through networks between individuals

To address issues such as asynchronous nature of functional module connections and operating information in unknown environments where communication delays and information loss can occur, we will develop data management mechanisms and data flow control functions on the network, establishing the base technologies for implementing network intelligence.
We will also achieve dynamic module allocation to each individual and enhance its robustness, taking into account physical distances between individual robots and swarms, network strength, communication status, and other factors.

Network intelligence RT platform

We will develop the base technologies to support the movements and tasks performed by robot swarms.
This platform will provide solid foundations for robots to operate cooperatively, enabling them to work in a variety of environments.
The RT platform will provide the infrastructure necessary for robot swarms to carry out high-level tasks, enabling stable activity in complex environments.

Integrated development of an RT platform with exploration, transportation and construction functions

To enable small worker robots to perform exploration, surveying and transportation tasks within lunar lava tubes, we will conduct R&D on functions such as integrated and collaborative tasking with transportation containers.
These containers will house and deploy payloads such as bases, communications and power relays, and they will be equipped with functions for the installation of various systems, functions to store and air-drop robot swarms, and cooperative functions supporting the informational and physical movement capabilities of the robot swarms.
This infrastructure will also be developed as a system integrated with the various equipment installed in the robots themselves.

Designing and implementing small RT jumping mechanisms

The small and lightweight robots to be developed will have a reduced ability to move over rough terrain considering their scale in comparison to the target environments.
To make up for this, jumping mechanisms that can be equipped in small robots will be developed as mechanical systems through research and development that takes into account energy consumption, the effects from terramechanics, and other factors.
We will also conduct research and development into methods for housing robots in containers and later deploying them, as well as functions to deliver containers and robots to sites with the use of air bags or other mechanisms.

Designing and implementing small RT surface movement mechanisms

We will conduct R&D into surface movement mechanisms that can be equipped in small robots that will operate over the rough terrain expected in lunar lava tubes, taking into account the perspective of terramechanics, power consumption, the volume and weight taken up by the mechanisms, and other factors.
We will implement this in small robots, and conduct R&D on methods for the storage and deployment of the robots into and out of containers as cooperative swarms. We will also study functions to deploy the robots on the lunar surface, and develop robot platforms to support lunar exploration missions.

Lunar lava tube exploration mission and exploration systems

Developing small exploratory robots adapted to the lunar environment based on exploration plans

We will determine the specifications of exploration systems in line with lunar mission plans, design those systems and conduct space environmental testing, and produce exploration systems that can be used in lunar missions.
The exploration systems will be made up of insertion capsules (containers) and the small exploration robots stowed within them. These will be loaded into the lander one capsule at a time.
Capsules placed on the lunar surface will deploy their payloads within a lava dome, and after releasing robots, will function as communication relay stations with the outside.

Expansion of Control Targets and Application Field for Network Intelligence System

We will expand the network intelligence system to control targets other than lunar and planetary exploration robots. For lunar and planetary exploration robots, we aim to robustly and continuously perform advanced tasks such as exploration by using AI to control multiple small robots of the same type. In particular, for the lower-level AI (swarm intelligence), we aim to stabilize the entire system by using a control method that forms a swarm without considering the location or control status of individual robots, which is caused by the low controllability and precision of small robots. We will apply this control method, which does not make the upper-level AI aware of individuals, not only to various robots but also to living organisms.

Collaboration Insect Cyborg with AI-Robot Technology and its Terrestrial Application

We will apply network intelligence systems to insects in particular. A key feature of the lower-layer AI is that it provides a control method that does not require the upper-layer AI to be aware of the individual. When this system is introduced into insects, its characteristics can be utilized to handle ambiguous data and perform robustly. Furthermore, we will consider collaboration schemes between this insect system and robot systems, and explore the possibility of developing applications, mainly on land, that are more closely related to people's lives.