1 Introduction In recent years, with the increasing demand for robots to replace or assist humans in power transmission line inspection, power transmission line inspection robots have become one of the research hotspots in the field of robotics both domestically and internationally.^1_2 Internationally, research on power transmission line inspection robot technology began in the 1980s. Representative research results include high-voltage line inspection robot prototypes developed by Tokyo Electric Power Company (TEPCO) and Mitsubishi Electric Corporation. TEPCO developed a prototype in 1989. This robot is mainly used for inspecting the outer steel wire and inner fiber aluminum film of optical fiber overhead ground wires. It has functions such as walking along optical fiber overhead ground wires, crossing obstacles, and inspection. The robot control system adopts motion control based on offline programming and precise positioning control based on sensor feedback information.^1_2 Domestic research began in the 20th century. In the 1990s, major research units included the Shenyang Institute of Automation of the Chinese Academy of Sciences, Wuhan University, the Institute of Automation of the Chinese Academy of Sciences, and Beijing University of Aeronautics and Astronautics. The inspection robot prototype of Wuhan University adopts a dual-arm structure and achieves the crossing of tower obstacles through remote control. The Institute of Automation, Chinese Academy of Sciences (CAS) proposed a three-arm inspection robot structure and verified its obstacle-crossing function through experiments. The Shenyang Institute of Automation, CAS, developed a prototype of an ultra-high voltage transmission line inspection robot and a prototype of an obstacle-crossing inspection robot with the ability to cross tower obstacles, and conducted experimental research. The obstacle-crossing inspection robot prototype from the Shenyang Institute of Automation, CAS, adopts a dual-wheel-arm composite mechanism, which can cross obstacles through various methods such as inchworm-like movement and single-arm rotation, depending on the shape of the obstacle and the operational requirements. In summary, inspection robots with obstacle-crossing functions are still in the development stage and require further improvement to meet practical application requirements. Research on ultra-high voltage transmission line inspection robots involves multiple technologies such as mechanisms, control, communication, and energy, among which the design of the control system plays a crucial role in realizing the overall performance of the machine. 2 Robot Mechanism and Kinematics Figure 1 shows a photograph of the obstacle-crossing inspection robot system with tower-crossing function developed by the Shenyang Institute of Automation, CAS. The system consists of a robot and a ground-based portable controller. The robot consists of two identical wheel-arm composite mechanisms and an inspection operation box. The wheel-arm composite mechanism is used to enable the robot to walk on overhead ground lines and cross pole obstacles. Its main components include walking wheels, gripping hands, and arms. The walking wheels and grippers are connected to the arms via a rotary joint and a telescopic joint, while the two arms are fixed to sliding rails. The robot body can slide on the sliding rails to adjust the robot's center of gravity to the gripping arm when the robot is gripping the line with one hand, thus maintaining the robot's posture stability. The robot body is mainly used to carry electronic equipment such as a gimbal camera, robot main controller, wireless transceiver, image transmission system, and motor driver. To achieve autonomous obstacle crossing, a kinematic model was established and the kinematic equations were derived. A simplified diagram of the inspection robot mechanism is shown in Figure 2. The robot grips the line with one hand and searches for the line with the other. Its kinematic equation is as follows: Where: d2 and d3 are the variables of the two rotary and two telescopic joints, respectively. Differentiating the above equation, the robot's Jacobian matrix equation can be obtained: Where: are the end-effector velocity of the gripper and the joint velocity of the robot, respectively. 3. Control System Design The operating environment of the power transmission line inspection robot is an ultra-high voltage power transmission line lightning protection line tens of meters above the ground. It is difficult for the operator to accurately judge the robot's operating status and issue corresponding control commands visually. Based on the inspection task and the practicality and reliability requirements of the field application, a control mode combining remote control and partial autonomy is adopted to enable the inspection robot to move along the line and cross obstacles. As shown in Figure 3, the inspection robot control system consists of two parts: the robot body control system and the ground remote control platform. The robot body control system adopts an embedded computer system based on the PC104 bus. Its components are as follows: core module MSM586SV, communication module EMM-8-XT, motion control module PMAC2A-PC104, A/D conversion module DMM-16-AT, and I/O module IR104. To realize the interaction of control information between the robot body controller and the remote control platform and the transmission of detection images, a wireless data transmission radio system based on serial i=1 communication and a microwave image transmission system are used respectively. The main functions of the robot's ground remote control platform are: to transmit control information and receive and store detection images via wireless communication with the robot itself. The main hardware components include: an IPC6008 main control computer, a microwave receiver, a PCI image acquisition card, and a wireless data transmission system. To meet the requirements of inspection operations, the key issues that the control system needs to address include: 1) the organic integration of remote control and local autonomous control modes; and 2) the realization of local autonomous obstacle-crossing control. This paper, considering the working characteristics of the inspection robot, establishes a finite state machine model of the inspection robot, realizing the organic integration of remote control and local autonomous control modes, and uses a combination of pre-programming and sensor positioning to complete the autonomous obstacle-crossing control of the inspection robot. 4. Software Design Based on Finite State Machine Model The main tasks of the power transmission line inspection robot include: 1) walking along the lightning protection line, adjusting the camera's observation angle and field of view according to the control information from the ground control console, stopping and collecting damage information after inspecting the line damage point; 2) when the inspection robot reaches the vicinity of the tower, it can overcome obstacles on the line (vibration dampers, single-point fittings, double-point fittings, crimped pipes, etc.) under the cooperative control of the operator and the robot controller, and then proceed to the next transmission line inspection. To facilitate the combination of autonomous and remote control of the robot, this paper establishes a finite state machine model of the inspection robot and implements the robot control software programming based on the finite state machine model. 4.1 Finite State Machine Model of the Inspection Robot The inspection robot is in a specific state at every moment during the inspection operation. Therefore, the control process of the inspection robot can be described by a finite state machine. The inspection robot's operation mainly includes six states, namely K = {waiting state, walking state, obstacle encounter state, obstacle crossing state, offline state, and error state}. The initial state is when the robot, having completed initialization, is placed on the overhead ground line and awaits control information from the remote control platform. The walking state involves the robot's wheels driving it along the line, with the camera and gimbal adjusting their posture according to remote control information to inspect the line. The detection state involves the inspection robot adjusting to a favorable position and posture after detecting a suspected fault point, and collecting image information of the suspected fault point. The obstacle encounter state is when the robot's environmental recognition system determines that the robot has moved near a pole, identifies the type and location of obstacles, and awaits remote control information. The obstacle crossing state involves the robot, under the coordinated control of remote control information and the robot controller, using the sequential movements of its two-wheeled arm composite system to cross the pole obstacle and enter the next span. The offline state is when the robot stops near the pole, waiting for personnel to remove it from the overhead ground line to complete the inspection work. The error state is when the inspection robot system malfunctions and cannot complete the inspection work normally. The inspection robot has two final states: F = {Offline State, Error State}. 4.2 Input and State Transition of the Finite State Machine Model. The operation process of the inspection robot can be described by the mutual transitions between states defined in the above state machine model. The inputs that trigger state transitions are the inputs to the finite state machine. The relationship between inputs and state transitions is described below. In the initial state, after receiving remote control information, the robot can enter the walking state. When the robot is in the walking state, different inputs result in different state transitions: when a suspected obstacle is detected, it transitions to the detection state; when the environmental recognition system detects an obstacle, it transitions to the obstacle encounter state; when a system error is detected, it transitions to the error state. After the robot enters the detection state, the operator can collect information on suspected fault points. After the operator completes the collection of suspected fault point information, the robot transitions back to the walking state. When the robot is in the obstacle encounter state, the operator's obstacle-crossing command can cause the robot to transition to the obstacle-crossing state or the offline state. In the offline state, the robot waits for the operator to remove it and does not perform any other state transitions. In the error state, the robot automatically transitions back to the walking state. Figure 4 illustrates the state transition relationship of the finite state machine model of the inspection robot. Figure 5 shows the flowchart of the robot control software. After initialization, the robot control program enters the initial state. Remote control commands and environmental information are used as external inputs to match the finite state machine model, and state transitions are completed. The control program performs operation control on the robot based on the remote control information. Establishing an automatic operation finite state machine model for the power transmission line inspection robot clarifies the robot's control process, which is beneficial for software programming and human-machine interaction. 5 Autonomous obstacle crossing control During the inspection operation, the robot sends images of the power transmission line and auxiliary equipment back to the remote control platform while walking along the overhead ground line. After completing the inspection task within one span (between two towers is called one span), the robot will cross obstacles such as towers and vibration dampers (see Figure 6) to enter the next span of the power transmission line. Since the inspection robot works at high altitudes, it is difficult for the operator to effectively remotely control it. Therefore, autonomously crossing tower obstacles is a control challenge for the inspection robot. This paper adopts a method combining pre-programming and sensor positioning to complete the autonomous obstacle crossing control of the inspection robot, taking into account the working characteristics of the inspection robot. As shown in Figure 7, based on the characteristics of the inspection robot mechanism described in this article, any obstacle-crossing process can be divided into two steps: 1) One robotic arm is fixed on the power line in front of the obstacle, while the other robotic arm moves around to the back of the obstacle via its joints; 2) The robotic arm accurately grasps the power line behind the obstacle through line-grabbing control. Repeating steps 1) and 2) allows the robot to cross the obstacle. To accurately determine the position of the power line during line-grabbing control, two laser sensors are installed on each robotic arm. The rotation of the robot's line-grabbing arm causes the laser beam emitted by one of the lasers in the line-finding arm to hit the power line, at which point the laser sensor generates a switching signal. After detecting this signal, the robot controller records the robot's current joint coordinates, the line-grabbing arm stops rotating, and the line-finding arm rotates instead, until the other laser sensor detects the power line. The robot's joint coordinates at this time are recorded, and the coordinates S1 and S2 when the two lasers detect the power line can be calculated from the recorded robot joint coordinate values. The position and orientation of the power transmission line can be determined by two points, S1 and S2. The line-finding control is transformed into controlling the beams of the two laser sensors to hit the straight line passing through S1 and S2. This control process can be decomposed into two simultaneous control steps: 1) Adjusting the position of the robot arm to be directly above the power transmission line; 2) Adjusting the orientation of the robot arm to be the same as the straight line passing through S1 and S2. Based on the above two steps, the following control algorithm can be designed: Theoretically, the above control algorithm can achieve accurate line grasping by the robot's gripper. However, due to mechanical structure gaps and processing errors, the robot's kinematic model has certain errors. To ensure the accuracy of line grasping, after completing the above control algorithm, the control result is detected by laser sensors. That is, the joint coordinates of the two laser sensors are recorded when they detect the power transmission line by rotating the line-finding arm. If the difference between the two joint coordinates is less than the allowable value, it is considered that the line has been accurately found; otherwise, the above control algorithm is applied to start the line-finding process again until the error is small enough. Autonomous obstacle-crossing control based on laser sensors can be accomplished through the following steps: 1) The grabbing arm rotates until one laser sensor detects the power line, then stops rotating and records the joint coordinates; 2) The finding arm rotates until another laser sensor detects the power line, then stops rotating and records the joint coordinates; 3) The robot is controlled to complete the finding of the power line based on the above two joint values; 4) Step 2) is repeated. If the difference between the two joint coordinates is less than the allowable error, the finding of the power line ends and the finding arm grabs the power line; if the difference between the two joint coordinates is greater than the allowable value, step 3) is repeated. 6. Experiments To verify the effectiveness of the inspection robot mechanism and control system, a simulated ultra-high voltage power transmission line was established in the laboratory. Obstacles included vibration dampers, suspension fittings, etc. Experimental research was conducted on the obstacle environment on the line. The robot body controller was programmed in C language, and the ground control terminal was programmed in VC++ language. During the experiment, the robot was first installed on the simulated line, and the robot program entered the initial state. After receiving the walking command from the remote control terminal, it entered the walking state. After the obstacle sensor detects an obstacle, the robot enters the obstacle-crossing state. The robot can autonomously overcome obstacles based on pre-programmed obstacle-crossing steps and laser sensor information, or enter a manual obstacle-crossing state according to operator instructions. Figure 9(a) shows the robot in walking mode, (b) and (c) show the robot crossing the vibration damper, and (d) to (f) show the robot crossing the suspension fittings. Experiments show that the robot can autonomously cross tower obstacles and perform multi-level inspection operations. 7 Conclusion and future work This paper introduces a power transmission line inspection robot that can autonomously cross tower obstacles and proposes the design and implementation method of the robot's control system. By establishing a finite state machine model of the inspection robot, the organic combination of remote control and local autonomous control modes is realized. The autonomous obstacle-crossing control of the inspection robot is completed by using a combination of pre-programming and sensor positioning. Experimental research was conducted in a laboratory environment simulating ultra-high voltage power transmission lines. The experimental results show that the robot can walk along the line and autonomously cross obstacles, thus verifying the effectiveness and rationality of the control system design. Future work will focus on improving the reliability of the robot in the complex environment of ultra-high voltage power transmission lines, solving electromagnetic compatibility issues, and conducting field experiments. This study analyzed and compared the characteristics of the Strider mechanism and the Flipper mechanism. Simulation results show that the Strider mechanism requires less adsorption torque and occupies less space during motion.