Research On Biomimetic Motion Control Of Self-Reconfigurable Robots In Unstructured Environments

Self-reconfigurable robot is a robotic system assembled by multiple simple modules with motor and sensory functions. It can change its configuration to dynamically adapt to the complex environments and diverse tasks through its connectivity and compatibility. Compared with the conventional robots, the self-reconfigurable robot has the features of self-reconfiguration, self-repair, self-adaptation, extensibility, and low-cost. Self-reconfigurable robot can perform outstanding advantages in the unstructured environments, such as emergency search and rescue, environment monitoring etc. However, the diverse configurations and multi-mode movement also make the motion control system much more complex than the one of the conventional robots. Therefore, this paper focuses on the unified control method of the self-reconfigurable robot. The control method can both control diverse configurations with multi-mode movement and adapt with the changes of the configuration. Self-reconfigurable robot can move automatically in the unstructured environments through adjusting its locomotion based on the external feedback.As the basic unit of the self-reconfigurable robot, the structure characteristics of the single module directly affect the mobility of the robots. Three generations of the self-reconfigurable robots have been designed and developed. The single module of the first generation is assembled by three orthotropic joint modules. The main research focuses on the independent movement of the single module. The single module can perform rectilinear motion, lateral motion, rolling motion, and turning motion. The second generation is developed for improving the accuracy in angular adjustment during the self-assembling process. The accuracy of the angular adjustment is increased to 2° through the proposed self-rotation mechanism. Both the structure characteristics of the first two generations are combined into the third generation. The proposed joint module consists of two orthotropic joints. The joint modules can be assembled to emulate PolyBot, iMobot, SuperBot etc. Towards the third generation, a homogeneous modular genderless connector is designed. The single modules can connect with each other in chain or in parallel. This connector improves the self-assembled and reconstructed function of the self-reconfigurable robot.In this paper, two joint modules of the third generation are assembled into a single module. Firstly, the influences of the joint status on the movement of robots are figured out through analyzing the features of the single module. A topological description method based on the adjacency matrix has been proposed. The adjacency matrix can fully describe the status of the single module, the connecting status, and the connecting orientations. This lays the groundwork for creating the locomotion- configuration library. The communication system of the self-reconfigurable robot has great impacts on the movement efficiency. A multi-level wireless communication system is proposed after comparing the current systems. The architecture of the system is composed of three levels, global communication network among robot, global communication network within robot, and local communication network among single modules. An adaptive transmission power adjusting method, network topology reconstruct method, and configuration identification method are proposed respectively based on the three network levels.Refering the lower nervous system of biological motion, a multilevel central pattern generator (CPG) based motion control system is established. The CPG nervous network consists of a rhythmic generator layer, a pattern generator layer, and a motor neuron layer. The interneurons and motor neurons are respectively modeled based on the Kuramoto phase oscillator and DMP. The impacts of each parameter on the output of the CPG nervous network are figured out through the single parameter analysis method. The coupling relations of the interneurons construct the topology of the CPG nervous network. Therefore, the impacts of the coupling relation and topological structure on the output of the CPG nervous network are analyzed. The configurations of the robots are divided into the none-legged type and the multi-legged type based on the structure feature and the motion mechanism. The corresponding CPG nervous networks are respectively established for the different robots. The I-shaped robot can perform travelling wave locomotion, serpenoid locomotion, sidewinding locomotion, and lateral rolling locomotion. The O-shaped robot can perform the rolling locomotion. Towards the multi-legged typed robots, the target functions of the motor neurons for each joint are established based on the foot trajectory planning. The H-shaped robot and the X-shaped robot can perform trot locomotion. Each configuration of the self-reconfigurable robot can adapt to different environments. It is found that the robots with similar structure can derive mutual movement when a simple reconfiguration or joint adjustment is performed. The derived movement can be performed when the phase relation of the interneurons and the target functions of the motor neurons are adjusted. And the topology of the CPG nervous network is maintained. The I-shaped robot and the O-shaped robot can derive the mutual movement. The H-shaped robot can derive the travelling wave locomotion of the I-shaped robot. It is shown that the proposed multi-level CPG nervous network can adapt to configuration changing of the self-reconfigurable robot.In order to make the self-reconfigurable robot adapt to the unstructured environments, the collision reflex, missing reflex, and vestibular reflex are proposed for the H-shaped robot. Therefore, the H-shaped robot can adapt to the rough terrain and slope. An autonomous obstacle avoidance method is proposed for the I-shaped robot. The continuity of the output of the CPG nervous network has great impacts on the stability of the robots. The continuity level of the output is discussed in the perspective of the curve continuity. A sigmoid based transition function is proposed for the parameter adjustment. It can ensure that the continuity of the output of the CPG nervous network is up to C3 which can avoid the damage to the joint motors.