Autonomous Solar Vehicle: A Community-Oriented Sustainable Transportation Solution Aligned with UN Sustainable Development Goals

Abstract

Introduction/Importance of Study: Urban transportation in developing countries presents persistent challenges, particularly for communities that rely on walking as their primary travel mode. Transport poverty affects millions of people and creates significant barriers to education, healthcare, and employment. In Pakistan, road-traffic accidents claim over 30,000 lives annually, of which approximately 67% are attributed to human error, while fossil-fuel dependency continues to exacerbate environmental degradation.

Novelty Statement: This research presents a novel community-oriented autonomous solar vehicle prototype integrating YOLOv8 Nano-based real-time detection with solar energy harvesting on a low-cost Raspberry Pi platform, providing safe, affordable, and sustainable last-mile transportation for universities, hospitals, and residential societies, directly contributing to UN SDGs 7, 11, and 13.

Materials and Methods: The prototype integrates a Raspberry Pi 4 Model B (4 GB) running ROS2 Foxy with YOLOv8 Nano (3.2 M parameters, 8.7 GFLOPs) for real-time pedestrian, vehicle, and obstacle detection. A custom dataset of 10,000+ campus-captured images was developed at Capital University of Science and Technology (CUST), Islamabad, and annotated using CVAT. Six ultrasonic sensors (four HC-SR04 and two waterproof JSN-SR04T) enable proximity detection over a 2–450 cm range, a NEO-6M GPS module provides waypoint navigation at 1 Hz, and a 50 W polycrystalline solar panel powers the 12 V system. The Pi is connected to a laptop external inference system through Wi-Fi or Ethernet over a TCP link on port 5555.

Results and Discussion: The model converged near epoch 85 of 100 (final box-loss ≈ 1.02, cls-loss ≈ 0.48, dfl-loss ≈ 1.18) and achieved an overall Detection Accuracy of 84.7% mAP@0.5 (per-class: pedestrian 87.3%, vehicle 84.1%, obstacle 82.6%) with 54.8% mAP@0.5:0.95. Inference speed was measured at 28 FPS on the companion laptop and 0.8 FPS on the Raspberry Pi 4, confirming the necessity of a two-tier compute architecture for real-time operation on low-cost edge hardware. The obstacle-avoidance subsystem achieved a 94% success rate across 50 controlled trials, with a total response time of 487 ± 42 ms (χ² = 3.18, p ≈ 0.037 versus an 85% baseline). Over a 14-day field test in Islamabad, the solar subsystem harvested 200 ± 15 Wh/day and avoided approximately 32 g CO₂/km relative to a grid-powered alternative, yielding a five-year per-vehicle cost advantage of approximately PKR 998k (≈ USD 3,560) over a fuel-powered equivalent.

Concluding Remarks: The prototype validates a cost-effective, community-focused transportation solution that addresses road safety through autonomous navigation while promoting environmental sustainability through solar energy integration, with particular relevance for educational campuses, healthcare facilities, and residential communities in developing regions.