Abstract: Due to increasingly competitive markets, the application scope of robots is expanding. However, each robot configuration can only adapt to a limited range, thus the flexibility of robots cannot meet the demands of market changes. The solution to this problem is to develop reconfigurable robot systems. This paper introduces the development status of reconfigurable robots and analyzes their research content and development direction. Keywords: Reconfigurability, Robot, Module 1 Introduction Theoretically, a robot is a flexible device that can be programmed to adapt to new tasks; however, this is rarely used in practical applications. Traditional robots are developed for specific applications. While these robots are sufficient for task-specific industrial applications, the increasing globalization of markets demands a wider range of robot applications. Each robot configuration can only adapt to a limited range, thus the flexibility of robots cannot meet the demands of market changes. The solution to this problem is to develop reconfigurable robot systems, which consist of a set of interchangeable modules with various sizes and performance characteristics, capable of being assembled into robots of different configurations to adapt to different tasks. Therefore, research on reconfigurable robot systems has attracted increasing interest from researchers and industrial applications. This paper analyzes the development status of reconfigurable modular robots and proposes directions for future research. 2. Domestic and International Research Status Extensive research has been conducted on reconfigurable robot systems abroad. Currently, the developed modular robot systems or reconfigurable robot systems mainly fall into two categories: dynamic reconfigurable robot systems and static reconfigurable robot systems. Dynamically reconfigurable robot systems include the metamorphic robot system developed by Pamecha and Chirikjian. It consists of a set of independent electromechanical modules, each with the ability to connect to, disconnect from, and traverse adjacent modules. Each module has a power source, allowing power and information input, which can be transmitted to adjacent modules. Configuration changes are achieved through the movement of each module on adjacent modules. This system possesses dynamic self-reconfiguration capabilities. Kotay et al. [21] proposed the concept of molecules (Mo[ecu[e]), where the modules of a self-reconfigurable robot are called molecules. Molecules are the foundation for building self-reconfigurable robots. Molecules connect with other molecules and can move on other molecules to form arbitrary three-dimensional structures, representing a dynamic self-reconfigurable system. A dynamically reconfigurable mobile robot was studied, which does not use wheels or tracks. Instead, it moves its center of gravity, i.e., the robot's movement, by moving modules called polygonal rod structures from the tail to the front, and can adapt to different environments through different configurations. Murata et al. proposed a three-dimensional self-reconfigurable structure. Its modules are homogeneous structures and there is only one type of module. Various structures are dynamically formed by the movement of one module on another. Static reconfigurable robot systems include: Benhabih0's modular robot, which proposed a modular robot unit based on remote drive technology. The drive method is similar to that of traditional industrial robots. It is believed that the drive part is too heavy, which affects the capability of the modular robot. Although this drive method reduces the flexibility of the modular robot, it is easy to implement and is a compromise solution. Paredis, Brown and Khosla Reconfigurable modular robot systems (RMMs) utilize a set of interchangeable links and joint modules of different sizes and characteristics. By assembling these general modules, various specialized robots can be assembled. This system is particularly suitable for reconfiguration and considers software reconfigurability. Chen et al. designed a modular reconfigurable robot, designing a module library and studying configuration design and kinematic and dynamic analysis methods. Han et al. developed a modular robotic arm, implementing configuration design through software for the mechanical design of modules. Hui et al. proposed an IRIS device, a modular, reconfigurable, and scalable robot system with two 4-DOF rotary joint robots, each reconfigurable into various configurations. Each joint is driven by a DC motor harmonic reduction gear and equipped with a position and torque sensor. Its software is also designed to be modular, scalable, and reconfigurable, just like the hardware. Fuiita et al. developed a reconfigurable robot platform based on Sony's OPEN. The R standard is used to establish various software and hardware modules, which are then used to assemble different robot structures. This platform is mainly used in the toy and entertainment industry. Matsumaru proposed the ToMMS system (Toshiba Modular Manipulator System), which consists of joint modules, link modules, and control units with joysticks. It allows for the creation of robots with various configurations by human intervention, and its kinematics are determined under specific configurations. Habibi et al. studied the design of reconfigurable hydraulically driven industrial robots. AMTEC in Germany produces POWERCUBE, a modular robot designed to meet various production needs with specific robots. Ji and Song proposed a design for a reconfigurable platform robot, focusing on the modular design of parallel robots. In terms of application scope, dynamically reconfigurable robot systems are mainly suitable for the toy industry and non-manufacturing industries, such as aerial robots and special robots for hazardous environments. Static reconfigurable robot systems are mainly suitable for industrial robots. 3. Design of Reconfigurable Modular Robot Systems A reconfigurable modular robot system is composed of a set of modules with different sizes and performance characteristics. These modules can be quickly assembled to form a robot best suited to complete a given task. Therefore, a reconfigurable modular robot system should have the following functions: (1) Users should be able to easily disassemble and assemble various modules to form different robot configurations to meet specific work requirements; (2) The number and types of modules used in the constructed robot configuration should be as small as possible; (3) Users should not have to make complicated modifications to the control software; (4) The assembled modular robot should be able to work immediately and complete the actual task. The main contents of the design of a reconfigurable modular robot system are the division of modules and the design of modules. The division of modules should take into account the application scope, workpiece characteristics and performance of the reconfigurable modular robot. At the same time, the modules themselves should also meet the following basic principles: (1) Each module unit should be independent and self-assembling; (2) Each module unit should be able to be quickly connected to any other module unit, regardless of its type; (3) Each module unit should have the minimum weight and minimum inertia; (4) Each module unit should be independent in kinematics and dynamics. Reconfigurable modular robot systems consist of a set of modules. Currently, mechanical modules are mainly classified into basic unit modules, end effector modules, link modules, and joint modules (mobile joint modules, rotary joint modules, and slewing joint modules). Several typical module classification methods include the robot module library established by Benhabib E et al., which divides modules into four categories: module unit connectors, link modules, main joint modules, and end effector modules. Link joints use a cylindrical shape with a circular cross-section to ensure resistance to bending and torsion in any direction, improving the flexibility of the constructed manipulator. The hollow structure ensures low module mass and rotational inertia. Basic components are also classified into this category. Main joint modules can be further divided into rotary joint modules and mobile joint modules. Based on their driving method, they are divided into R-Actuator-M modules (used for joints close to the basic components; it uses an I²C motor with harmonic reduction drive and is heavy), R-Actuator1, and Link M. (For joints far from the base, power is not placed on the joint itself, but transmitted to the joint via a transmission element closer to the base), P—Actuator-M (DC motor driven kinetic joint). End-effector joints are also divided into rotary joints and kinetic joints, R—Actuator-E and R—Actuator, Link-E, whose driving principle is the same as that of the main joint module. A unit connector was also designed, using a stop and locating pin for positioning and bolts for connection. ParedisE263 et al. proposed a hardware module division method in the RMMS system, establishing the basic modules of the manipulator: a link module, three rotary joint modules, and one slewing joint module. The basic module and link module are without degrees of freedom, while the joint module has one degree of freedom. The modules are self-contained, and the hardware includes a CPU, sensors, drivers, brakes, and transmission devices. The system includes components such as sensor interfaces, motor amplifiers, and communication interfaces. The electrical components are also designed based on modular principles, using a motherboard with basic functions and daughterboards with special functions. An integrated quick-coupler is also designed for rapid connection between modules. In the modular robot system designed by Chen et al., only link modules and joint modules are considered. The joint modules include rotary joint modules, kinetic joint modules, helical joint modules, and cylindrical joint modules. Link modules are designed as cubes and cuboids, characterized by multi-joint connections and geometric symmetry. All six surfaces of the cube have connection interfaces, and the two ends of the cuboid also have connection interfaces. The basic functions of a reconfigurable modular robot system module should include the following aspects: (1) The module should have self-encapsulation function to complete a specific function. (2) The module should have driving capability to complete specific motions and actions. (3) The module should have communication capability so that the modules can work in coordination. (4) The module should have data processing capability. 4 Configuration Design of Reconfigurable Modular Robots The reconfigurable modular robot system is composed of a set of modules with various functions. Different modular robots can be assembled by selecting different module combinations. The purpose of reconfigurable modular robot configuration design is to find an optimal assembly configuration to complete the given work. The method of reconfigurable modular robot configuration design mainly considers the following three issues: First, determine the expression method of the configuration; second, determine the evaluation criteria of the configuration; and finally, use an appropriate optimization method to determine the optimal configuration that satisfies the given task. Currently, the expression methods of the configuration are: Chenis et al. used the concept of graph theory to express the assembly relationship of the modular robot using the assembly correlation matrix and established the assembly configuration evaluation function: is the structural performance of the modular robot assembly configuration. A is an assembly correlation matrix. The performance measurement value of the modular robot in the worst case of the task point set is expressed as the task evaluation standard. A genetic algorithm is used to solve this optimization problem to obtain the optimal robot configuration. Paredis et al., based on the requirements of the kinematic design task (i.e., accessibility, joint movement within the limit range, and obstacle avoidance requirements), only considered rotational joints and solved the problem of kinematic iteration judgment through substitution in the kinematics expressed by D-H parameters, and used simulated annealing to optimize the configuration. ChocronEo3 and Hanr" used a genetic algorithm for modular robot configuration design. 5. Kinematics and Dynamics of Reconfigurable Modular Robots The generation of kinematics and dynamics of reconfigurable modular robots also differs from that of traditional robots, requiring automatic generation of kinematics and dynamics after configuration changes. Benhabib et al. proposed a motion module technology, using D-H parameters, where the motion connection between the input and output ports of each mechanical module is transformed through a 4×4 homogeneous transformation matrix. Chenn et al. used an exponential product approach in the kinematic analysis of modular robots. They first studied the kinematics of a connecting pair (consisting of two links and a joint module) and expressed it as an exponential product. Then, based on the robot's configuration, they used a series of exponential products to obtain the robot's forward kinematics. 6. Control System Software for Reconfigurable Modular Robots Reconfigurable robots require both hardware reconfigurability and real-time control software reconfigurability to adapt to rapid changes in the robot's application scope. To achieve software reconfigurability, it is necessary to solve the problems of developing reconfigurable software module libraries and automatically integrating these modules according to hardware configuration and control tasks. In the field of computer software engineering, there has been extensive research on software reuse, but research on reconfigurable software is relatively limited. Research on the reconfigurability of control software for reconfigurable robots is also scarce. Khosla et al. designed a dynamically reconfigurable real-time software and a multi-layered human-machine interface based on a sensor-based real-time system using port-based objects and applied it to a reconfigurable modular robot. 7 Research Directions for Reconfigurable Modular Robots In summary, the research on reconfigurable modular robot systems has attracted the attention of many researchers. Currently, further research is needed in the following areas: (1) Functional design and implementation methods of modules in reconfigurable modular robot systems. This includes robot function analysis and allocation, software and hardware function analysis of modules, research on module description methods, design of software and hardware modules, and research on automatic or rapid connection methods for software and hardware modules. (2) Configuration design of reconfigurable modular robots. This includes research on methods for describing the tasks required by the robot, research on robot configuration expression methods, and research on methods for generating optimal robot configurations. (3) The kinematics and dynamics research of reconfigurable modular robots should primarily consider the reconfigurability of the software. This includes analysis methods for module kinematics and dynamics, and research on kinematics and dynamics analysis methods for distributed modular robots. (4) Research on reconfigurable real-time control software suitable for reconfigurable modular robot systems. This includes research on functional analysis and partitioning methods of robot control modules, and research on software reconfiguration methods.