Autonomous Inspection Robot for 110kV Transmission Lines: Design and Implementation of a Three-Arm Suspended System

Abstract​

To address the inherent limitations of manual inspection and aerial surveying for high‑voltage transmission lines, this proposal introduces an autonomous inspection robot specifically designed for 110 kV power lines. Featuring an innovative three‑arm suspended mechanical structure, the robot integrates autonomous crawling, obstacle negotiation, online power harvesting, and multi‑fault diagnosis. It aims to automate and intellectualize line inspection, significantly improving both the efficiency and safety of grid operation and maintenance while reducing costs.

​I. Project Background and Objectives​

​1.1 Background: Challenges of Traditional Inspection Methods​

High‑voltage transmission lines, being continuously exposed to outdoor environments, are prone to defects such as broken strands and wear due to mechanical tension, electrical flashover, and material aging, thus requiring regular inspection. Current methods face significant bottlenecks:

• ​Manual Inspection:​​ Labor‑intensive, inefficient, high‑risk, and heavily constrained by weather and terrain.
• ​Drone Aerial Surveying:​​ High operational cost, limited endurance, subject to airspace control and adverse weather, and challenging for close‑range defect detection.

​1.2 Objectives: An Intelligent Inspection Alternative​

This project aims to develop an autonomous inspection robot for 110 kV high‑voltage transmission lines capable of replacing manual labor. Core objectives include:

• ​Functional Autonomy:​​ Achieve autonomous crawling and precise obstacle negotiation (e.g., crossing vibration dampers and clamps) on the lines.
• ​Intelligent Detection:​​ Integrate visual and infrared sensors to automatically identify and diagnose typical faults such as broken strands.
• ​Energy Self‑Sufficiency:​​ Utilize non‑contact inductive power harvesting technology for online self‑replenishment, enabling long‑distance inspection.
• ​Maximized Efficiency:​​ Greatly enhance inspection efficiency and data accuracy, thereby reducing operational costs and safety risks.

​II. Core Technical Solutions​

​2.1 Innovative Mechanical Structure Design: High Mobility and Stability​

• ​Overall Structure:​​ Adopts a three‑arm suspended configuration that combines the advantages of multi‑segment separated and wheel‑arm composite mechanisms, balancing the efficiency of wheeled movement with the stability of inchworm‑like creeping. Total weight is approximately 29 kg.
• ​Key Components:​​
• ​Flexible Arms:​​ The front and rear arms employ a double four‑bar linkage mechanism, driven by a total of 16 motors, allowing independent or coordinated pitch motion with joint stiffness‑flexibility smooth transition capability to adapt to complex line conditions.
• ​Drive Unit:​​ Uses high‑power Swiss Maxon DC motors with center‑separated drive wheels, providing strong obstacle‑crossing ability (capable of passing vibration dampers) and gradeability (routine 60°, up to 80° with braking).
• ​Braking Unit:​​ Employs a spiral‑crank slider self‑locking mechanism to effectively prevent accidental slipping or falling during slope traversal or obstacle negotiation.
• ​Kinematic Validation:​​ Inverse kinematics analysis based on the CCD iterative algorithm; simulations show convergence in only 7 iterations, efficiently validating the robot’s ability to achieve complex poses such as crossing suspension clamps and 45° turn jumpers.

​2.2 Hierarchical Intelligent Control System: Seamless Autonomy and Remote Control​

• ​System Architecture:​​ Adopts a three‑layer distributed control structure (upper ground management layer, middle robot planning layer, lower execution layer), coordinated by a PC/104 industrial computer and an ATmega128AU microcontroller for real‑time decision‑making and execution.
• ​Hybrid Control Strategy:​​
• ​Autonomous Mode:​​ Offline path planning based on a pre‑set knowledge base, combined with real‑time sensor feedback for fully autonomous crawling and obstacle negotiation.
• ​Remote Control Mode:​​ In extremely complex environments, ground operators can perform joint‑level fine manipulation or issue macro‑commands via remote intervention, supported by HD video (25–30 Hz) transmitted from the robot.
• ​Performance Metrics:​​ Single inspection distance ≥ 2 km, average speed ≥ 0.9 m/h, image transmission distance ≥ 2 km.

​2.3 Online Inductive Power Harvesting & Intelligent Power Management: Unlimited Endurance​

• ​Power Harvesting Principle:​​ Uses a split‑core current transformer to inductively harvest energy from the magnetic field around the high‑voltage conductor. The CT core is made of high‑permeability iron‑based nanocrystalline alloy; an optimized design enables a low starting current of 32 A.
• ​Power System:​​ Delivers stable rectified voltage; output power covers a line current range of 32 A to 10 kA. Equipped with a 24 V/12 A·h intelligent Li‑ion battery pack using a three‑stage charging algorithm, with over‑temperature protection for safety, efficiency, and long service life.

​2.4 Machine Vision Obstacle Recognition: Accurate Navigation​

• ​Recognition Targets:​​ Accurately identifies key obstacles such as suspension clamps, straight‑line jumper clamps, and turn jumper clamps.
• ​Algorithm Flow:​​
• ​Positioning:​​ Coarse positioning via sub‑block grayscale analysis, precise identification of the transmission line via histogram equalization and threshold segmentation.
• ​Feature Extraction:​​ Extracts obstacle contours using morphological operations, analyzing left/right edge slopes as classification features.
• ​Recognition:​​ Applies a fuzzy pattern recognition algorithm based on the maximum membership principle for fast and accurate obstacle type identification.
• ​Performance:​​ Single image processing time ≈ 108 ms; reliably identifies typical obstacles, providing real‑time input for obstacle‑negotiation decisions.

​2.5 Broken Strand Intelligent Diagnosis: Accurate Fault Warning​

• ​Detection Principle:​​ Based on the phenomenon of localized resistance increase and temperature rise due to broken strands, uses an infrared sensor to detect thermal radiation signals.
• ​Intelligent Diagnosis Model:​​
• ​Signal Processing:​​ Uses the db4 wavelet base for 6‑layer decomposition to filter out noise and focus on frequency bands containing fault features.
• ​Feature Extraction:​​ Introduces wavelet energy entropy to characterize signal complexity, combined with peak‑to‑peak values of detail components, forming a 4‑dimensional feature vector.
• ​Diagnosis Decision:​​ Uses a 3‑layer BP neural network for diagnosis. Experimental verification shows 100 % accuracy on test samples and a 98 % online detection success rate.

​III. Solution Advantages Summary​

• ​High Adaptability:​​ Three‑arm flexible structure provides excellent obstacle negotiation and terrain adaptability.
• ​High Autonomy:​​ Hybrid control system enables long‑distance autonomous inspection with remote intervention capability.
• ​Long Endurance:​​ Innovative online power harvesting fundamentally solves endurance limitations.
• ​Accurate Detection:​​ Integration of machine vision and infrared thermography with intelligent algorithms ensures high fault‑recognition accuracy.
• ​Safe and Economical:​​ Replaces high‑risk manual work, reducing safety hazards and long‑term operational costs.

​IV. Current Limitations and Future Prospects​

​4.1 Current Limitations​

• Still requires minimal manual assistance in extremely complex line environments.
• Potential for further optimization of mechanism size and obstacle‑negotiation stroke for a more compact design.
• Power system starting current remains relatively high, limiting application on very low‑load lines.
• Current fault detection types are mainly focused on broken strands; the range of detectable faults can be expanded.

​4.2 Future Outlook​

• Mechanism lightweighting and balance optimization to improve obstacle‑negotiation efficiency and stability.
• Integration of multi‑sensor navigation to enhance positioning and environmental perception accuracy.
• Optimization of the power harvesting circuit to further reduce the starting current and expand the application range.
• Expansion of the fault diagnosis library to include defects such as damaged insulators and contamination.
• Enhancement of the robot’s reliability, improving industrial‑grade protection (e.g., dustproof, waterproof, and EMC capabilities).