Advancing Autonomous Navigation with Nav2
I've been working on improving the autonomous navigation capabilities of Store-Picker-Bot, focusing on integrating ROS 2 Nav2 to make the robot more capable of navigating complex, dynamic environments.
What I've Been Working On
- Nav2 Navigation Improvements: Fine-tuning navigation behavior to improve goal reaching, turning, and maneuvering through narrow and congested spaces.
- Dynamic Obstacle Avoidance: Improving the robot's ability to navigate around moving obstacles while maintaining a safe clearance from surrounding objects.
- RL-Based Navigation: Developing and testing an RL navigation policy to improve decision-making in environments containing both static and moving obstacles.
- Real-World Robot Testing:Validating navigation behavior on the physical robot and refining controller parameters based on observed performance.
One of the key challenges is balancing obstacle clearance, efficient path planning, and reliable goal reaching in dynamic environments. I'm working to make navigation smoother and more robust without unnecessarily increasing the path length or slowing the robot down.
The next step is to continue refining the Nav2 and RL navigation pipeline for the current robot configuration and strengthen its integration with the overall Store-Picker-Bot workflow.
The goal is to make the robot navigate reliably between store locations so it can eventually support the complete process of locating shelves, collecting products with the SO100 robotic arm, and bringing them to the packing area.
Building toward more intelligent, reliable, and adaptive autonomous navigation with Physical AI