RESOLV

Task Definition: Autonomous Solar Panel Maintenance

"Design and demonstrate an autonomous robot that can inspect solar panels, detect dust or sand accumulation on a solar panel or other defects, and independently initiate a cleaning action without human intervention making the best decision based on the resource constrained settings in space."

• Detect surface contamination using computer vision and color blocking
• Autonomously determine when cleaning is required
• Perform a mechanical cleaning action using onboard actuators
• Operate continuously without manual control

Inspiration

With solar power being a cornerstone of intergalactic travel and settlement, we were inspired to create an autonomous system capable of cleaning and monitoring solar panels without any human intervention.

Our goal was to design a robot that could maintain power systems reliably in harsh, remote environments, like lunar or Martian solar farms, where sending humans for routine maintenance is costly, slow, and dangerous. Beyond just cleaning, we envisioned a framework with the potential to detect not only dust but also physical damage, such as holes or scratches, revolutionizing the way intergalactic energy production is maintained and monitored.

What it does

RESOLV autonomously cleans and maintains solar panels. It detects dust or, in our demonstration, sand on the panels and moves a cleaning mechanism across the surface.

The robot maps its environment using LIDAR and a live camera feed, allowing it to avoid obstacles and track its position in real-time.

All collected data, including the camera feed, cleaning status, and sensor readings, is displayed on the front-end live, providing complete visibility into the robot’s operation.

How we built it

• Hardware: We created a fully custom 3D-printed chassis design with TPU-compliant wheels, as well as TPU springs used to create an all-terrain suspension drive system. We also have an articulated servo arm and an in-house designed lidar system with an ultrasonic sensor and a servo.

• Electronics: Our motors are controlled via an L298N motor driver, while the servos are managed through a PCA9685 16-channel servo driver. The Raspberry Pi 4B runs a multi-threaded system, allowing the robot to collect sensor data, control movement, and transmit information simultaneously. Wires are connected through a perf board and use ferrules and screw terminals to maintain repairability and modularity.

• Software: The backend software autonomously controls the robot, using the camera feed to detect dust on solar panels and determine where to clean. It aligns the cleaning servo over dirty areas, tracks which panels have been visited, and adjusts the robot’s movement to center the panel for cleaning. The software coordinates cleaning cycles, switching between motion and cleaning actions, and continuously updates sensor readings and the LIDAR-based environment map. By managing these tasks simultaneously, the system ensures fully autonomous operation across multiple panels, even in complex or cluttered environments.

• Front-end: The frontend is an interactive, real-time dashboard built using Streamlit, providing a complete overview of the robot’s operation and environment. It streams a live camera feed for monitoring the solar panels, alongside a processed vision feed that highlights detected dust or sand. The dashboard also displays a LIDAR radar map to visualize the surroundings, helping the robot avoid obstacles and track its position. Real-time servo telemetry shows the angles of the cleaning and LIDAR servos, while the cleaning status indicates which areas of the panel are being serviced. This interface combines live operational feedback and sensor visualization in a user-friendly dashboard, giving an intuitive view of RESOLV’s autonomous functionality.

Challenges we ran into

• Mapping the solar panel surface and sand location accurately using CV and color-mapping
• Achieving smooth cleaning motion while analyzing
• Integrating multiple sensors and real-time data visualization for the front-end
  

Accomplishments that we're proud of

• Fully autonomous dust-detection and cleaning operation showcased in our demo
• Real-time sensor integration with live mapping on the front-end
  
• Reliable active-detection of quantity and location of dust on solar panels using CV

What we learned

• How to integrate sensors and control systems in a real-world robot.
  
• The importance of real-time feedback for autonomous decision-making.
  
• Challenges of robotics in remote and delicate environments, like solar panels.

What's next for RESOLV

• Testing on larger solar arrays and refining navigation on uneven surfaces.
• Adding more advanced environmental sensing to handle sand or debris in outdoor conditions
• Improving cleaning efficiency and battery management for extended autonomous operation -Implementing an integrated machine model to react critically to problems (such as dents, holes, or scratches) on the solar panel and resolve the issues -Develop frameworks enabling agents to make trustworthy decisions when conducting repairs in space -Apply this decision matrix or decision system to other applications, including healthcare in space, construction in space, as the inputs and decisions needed to be made by the model can be tuned.

Built With

• opencv
• python
• raspberry-pi
• streamlit