Mobile Quadruped Robot Navigation

During my internship at AlphaZ, I was responsible for collaborating with Unitree Robotics and Intellif to develop an indoor and outdoor mobile navigation system based on the Unitree B2 quadruped robot. The project aimed to enhance the autonomous navigation capabilities of quadruped robots in complex environments.

Project Objectives

The main objectives of this internship project included:

• Designing a robust navigation system for the quadruped robot
• Designing a state estimator for the robot
• Implementing terrain analysis, obstacle detection and avoidance in complex environments
• Optimizing motion control and path planning for the quadruped robot
• Implementing a human-machine interaction remote control interface and remote communication

Technical Approach

Our technical approach involved several key components:

1. Using ROS as the robot operating system framework
2. Environment perception based on LiDAR and cameras, utilizing gridmap
3. Improved motion control and local obstacle avoidance based on PD and sampling-based control algorithms
4. Utilized combination of A* and Graph-based global planning

Implementation Results

Dynamic Obstacle Avoidance

Climbing Stairs Autonomously

Key Achievements

The main achievements of this internship project include:

• Successfully implemented indoor and outdoor autonomous navigation for the quadruped robot from scratch
  • Achieved pose observation and odometry in oscillating environments
  • Enhanced obstacle avoidance response speed while analyzing terrain passability
  • Achieved continuous patrolling for a battery cycle (3-4 hours), with a near 100% success rate for complete cycle patrols, excluding failures due to hardware issues
• Developed an intuitive user interface and network interface for easy operation and control
  • Deployed a WebUI for robot control
  • Integrated into a multi-robot intranet system, enabling remote video monitoring and remote takeover in a cellular environment

Impact and Gains

Through this internship, I gained valuable experience in quadruped robot control and navigation.

Additionally, during the tuning and exhibition process, I realized that for general-purpose robots to truly achieve commercialization, the focus should be on perception and interaction capabilities.