Neurealm developed a high-performance Autonomous Mobile Robot (AMR) platform built on NVIDIA AGX and TI Jacinto 7 hardware. The system features a robust Docker-based ROS environment integrated into a CI/CD workflow. By utilizing 2D LiDAR and Intel RealSense D435i cameras, the platform achieves precise mapping, localization, and dynamic obstacle avoidance.

Business Context

The project focused on creating a scalable, reliable autonomous platform for industrial and commercial use. Key drivers included:

• Advanced Platforms: Integration with edge-AI hardware (NVIDIA AGX, TI Jacinto 7).
• Operational Agility: Achieving speeds of 0.5 m/s, rotational speeds of 180°/s, and a 3 kg payload capacity.
• Modern DevOps: Utilizing Docker-based ROS setups to streamline development and deployment.
• Navigation Reliability: A requirement for high-precision path planning and sensor fusion in complex environments.

Solutions

Neurealm implemented a comprehensive navigation and perception stack:

• Mapping & Localization: Employed 2D LiDAR SLAM (Gmapping/Hector SLAM) and Particle Filter-based localization.
• Autonomous Navigation: Utilized Probabilistic Path Planning (A* with Local Dynamic Window Approach) and Dynamic Obstacle Avoidance.
• Sensor Fusion: Integrated IMU, Wheel Odometry, and 2D LiDAR for precise odometry.
• Control & Perception: Integrated D435i cameras for environmental awareness and enabled WiFi-based teleoperation for manual steering and throttle control.