PROJECT TOPICComputer Vision & Embedded AI
Real-time Spatial SLAM Rover
An autonomous robotics project concept exploring visual-inertial odometry, 3D point cloud generation, and octree map representation running entirely on an onboard embedded GPU processor.
Autonomous Spatial Awareness & Navigation
- ✓Integrates multi-sensor fusion combining 2D LiDAR, stereo depth, and IMU data.
- ✓Explores embedded GPU acceleration for real-time feature extraction and loop closure.
- ✓Constructs volumetric 3D occupancy grid maps for complex multi-level obstacle avoidance.
The Engineering Challenge
Problem / Objective
GPS is unavailable indoors, requiring autonomous mobile robots to construct spatial maps and localize simultaneously using sensor fusion.
Conceptual Signal Flow
System Concept
RGB-D Camera + LiDAR → Visual Odometry & EKF → 3D Octomap → Path Planner.
Architecture Modules
1Differential Drive Rover Platform
2Stereo RGB-D Vision Sensor
3360-Degree 2D LiDAR Unit
4Embedded AI Compute Module (Jetson)
Hardware Categories
Nvidia Jetson / Raspberry Pi 5Depth Camera (RealSense)RPLiDAR A2Wheel Encoders
Technologies & Software
ROS 2RTAB-MapVisual OdometryPython/C++CUDAROS 2 Humble / IronNav2 StackCartographer / RTAB-Map
Engineering Considerations & Edge Cases
- Processing latency and thermal throttling under dense point cloud generation
- Wheel odometry slip compensation on smooth floors
Applications
Indoor mapping & surveillance
Hospital delivery rovers
Autonomous inventory auditing
Engineering Inquiry & Collaboration
Interested in Developing this Architecture?
Connect directly with Tamizh Tech engineers in Coimbatore to discuss mechanical fabrication, sensor selection, firmware implementation, or turnkey system commissioning.