Organization: SRM University, India
Timeline: Sep 2023 – Aug 2024

Robotic Manipulation · Computer Vision · YOLOv8 · Mechanical Design · Ansys · Patent · Q1 Journal Publication

Problem Context

Manual EV charging can be inconvenient or inaccessible in constrained parking environments, particularly for users with limited mobility. Automating the charging process requires precise perception, reliable manipulation, and mechanical robustness across different vehicle geometries and connector standards.

This project focused on developing an autonomous robotic arm capable of detecting EV charging ports and performing accurate, hands-free docking across multiple connector types.

System Overview

The system combines a custom-designed 6-DOF robotic manipulator with vision-based charger detection to enable autonomous alignment and docking. Camera input is processed using a deep-learning-based object detection pipeline to localize charging ports, while the robotic arm executes controlled motion to position and insert the connector.

Mechanical design, perception and control were integrated into a single system to validate end-to-end autonomous charging capability under realistic operating conditions.

My Role & Responsibilities

Role: Robotics system design and perception lead

• Designed the full 6-DOF robotic manipulator in SolidWorks with emphasis on reach, workspace coverage, and torque balance.

• Performed structural and modal analysis in Ansys to ensure mechanical stability, fatigue resistance, and safe operation.

• Developed vision-based charging port detection using YOLOv8 across multiple connector standards
  (Type-1, CHAdeMO, GB-T, Tesla).

Key Technical Decisions

• Chose a custom manipulator design to accommodate diverse vehicle geometries and charging port locations.

• Used learning-based visual detection to generalize across connector types rather than relying on fixed
  geometric assumptions.

• Evaluated detection and docking performance separately to identify perception-driven versus mechanical
  sources of error.

Results & Impact

• Achieved >92% successful docking accuracy across multiple charging standards.

• Obtained mAP50 of 0.993 for Type-1 connectors and 100% recall for Tesla charging ports.

• Demonstrated improved safety, accessibility, and repeatability for autonomous EV charging applications.

• Work resulted in a filed Indian patent and a Q1 Elsevier journal publication (Results in Engineering, Vol. 26, 2025).

Learnings & Limitations

• Precise mechanical alignment is as critical as perception accuracy for reliable autonomous docking.

• Vision-only perception is sensitive to lighting variation and partial occlusion around charging ports.

• Future improvements could include force sensing or compliance control to further increase robustness during connector insertion.

Maximum Stress & Deformation in all 6 joints with different material

Comparison of Joint Moments with different Payloads