چكيده
Quadruped robots, due to their unique abilities in interacting with complex environments,
are recognized as outstanding options for advanced navigation systems. Their agile and dynamic
design enables movement along difficult paths, where wheeled robots face limitations. The
ability of these robots to overcome large obstacles and move in various directions makes them
an ideal choice for navigating rough terrains. On the other hand, a robust and efficient vision
system can help robots accurately perceive their surroundings. Given these capabilities and the
importance of navigation as a key indicators for autonomous robots, this research focuses on
developing and implementing an advanced navigation system for the quadruped robot at the
Iran University of Science and Technology.
In this project, the main goal was to develop a navigation system for the quadruped robot that
can autonomously move in complex environments. Inspired by well-known quadruped robots
such as ANYmal and Mini-Cheetah, we aimed to follow similar navigation and path planning
development processes of these robots. Considering the successful experiences of these robots
in complex environments, we also implemented similar algorithms and drew inspiration from
their agile and dynamic designs. To achieve this, vision and motion sensors, including the
RealSense camera and IMU, were used to collect environmental data and construct elevation
maps. Using the TEB and A* planning algorithms, the robot was able to identify and follow
optimal paths in various environments. An important aspect was ensuring that the system
operates in real-time, allowing the robot to continuously and promptly make appropriate decisions
when encountering obstacles and paths. The system was first implemented in a simulation, on
a Core i7 processor and then on Jetson Nano. With proper system settings and optimization of
system rates, the robot successfully moved in unknown environments and overcame obstacles.
This project not only developed navigation algorithms but also examined the challenges of
implementing them in real environments with limited hardware
