Low-Cost Detection and Tracking of Dynamic Objects for Autonomous Mobile Robots

This research develops cost-effective solutions for dynamic object detection and tracking using affordable sensors and computational resources, addressing critical challenges of hardware constraints, algorithm efficiency, and real-time performance requirements. The work creates lightweight deep learning models including MobileNet and EfficientNet architectures, quantized neural networks for reduced memory usage, and efficient tracking algorithms with Kalman filtering, correlation filters, and particle filters with reduced computational complexity. Key technical innovations include multi-threaded processing frameworks, memory-optimized data structures, visual-inertial sensor fusion, adaptive processing based on available resources, and energy-efficient operation with dynamic voltage scaling and power management strategies for extended autonomous operation.

Service robotics and assistive technology applications benefit from democratized access to autonomous capabilities including home cleaning robots, delivery systems, navigation aids for visually impaired individuals, and educational robotics platforms. The work demonstrates significant cost reduction through hardware optimization, component selection strategies, and open-source software frameworks while maintaining acceptable detection accuracy and real-time performance. Research communities and robotics companies can leverage this expertise for developing affordable automation solutions, expanding educational accessibility, enabling developing country deployment opportunities, creating personal robotics applications, and advancing edge AI technologies that reduce barriers to entry for small businesses and community-driven robotics development initiatives.

Acknowledgments

We acknowledge support from affordable robotics research initiatives and open-source hardware communities. This work was conducted with focus on accessibility and cost-effectiveness for global deployment.

---

Note: Full content extraction from PDF required for complete details. Please refer to the original publication: 23-ral-zhefan-xu-and-xiaoyang-zhan-low-cost-detection-and-tracking-of-dynamic-objects.pdf