This project presents a few-shot dynamic obstacle avoidance technique for mobile robots in unknown environments, leveraging advancements in deep learning algorithms and mobile platforms[^2^][2].
- NVIDIA JetBot: An AI kit with a NVIDIA Jetson Nano-based system[^3^][3].
- Convolutional Neural Networks (CNNs): Utilized for training in a Reinforcement-Learning fashion via PyTorch[^4^][4].
- Linux4Tegra OS: The operating system running on Jetson Nano[^5^][5].
- Jetpack SDK: Includes tools and libraries for development on Jetson Nano.
The project involves training a CNN to enable a self-driving robot (JetBot) to navigate a fictitious road autonomously[^6^][6]. The model is trained using PyTorch and deployed on the mobile robot for obstacle avoidance recognition tests[^7^][7].
The tests conducted on the mobile robot platform showed good performance, validating the proposed strategy[^8^][8]. The project also discusses the potential applications and limitations of the approach.
Further research will focus on enabling the robot to retain information acquired in various testing contexts and improve its adaptability to different environments.
- A comprehensive list of references is included in the project documentation, providing a foundation for the research and development of the obstacle avoidance strategy.
Contributions to the project are welcome. Please refer to the CONTRIBUTING.md file for guidelines on how to make contributions.
This project is licensed under the MIT License - see the LICENSE.md file for details.
Special thanks to the Department of Electronics and Communication Engineering, Nirma University, for their support and guidance throughout the project.