Mobile Robotics Gallery

TerraMax - Autonomous Agricultural Robot

Source: EarthSense | Website: TerraMax Product Page

TerraMax represents the cutting edge of autonomous agricultural robotics, designed to revolutionize precision farming through advanced sensor fusion and autonomous navigation capabilities. This system demonstrates real-world applications of mobile robotics concepts covered in CS498GC.

Key Technologies Demonstrated:

• Autonomous Navigation: GPS-based path planning and obstacle avoidance
• Sensor Fusion: Multi-modal sensor integration for crop monitoring
• Computer Vision: Plant detection and health assessment
• Precision Agriculture: Targeted intervention and data collection
• Robotic Mobility: All-terrain autonomous operation

Course Relevance:

• Modules 1-2: Coordinate transformations in agricultural mapping
• Module 4: Sensor fusion for crop monitoring
• Module 6: Localization in outdoor environments
• Module 7: SLAM applications in precision agriculture

The TerraMax demo video that was embedded here in Fall 2025 is no longer available on YouTube.

Mobile Manipulator for Active Semantic Mapping

Paper: J. Cuaran, K. S. Ahluwalia, K. Koe, N. K. Uppalapati, G. Chowdhary, Active Semantic Mapping with Mobile Manipulator in Horticultural Environments, ICRA 2025 (arXiv 2412.10515)

A mobile robot manipulator with an RGB-D camera builds probabilistic semantic maps of horticultural crops and chooses where to look next: it detects semantic targets, generates candidate viewpoints and scores them by information gain. The paper reports real-world experiments, which also show challenges such as depth sensor noise and changing environmental conditions.

Key Technologies Demonstrated:

• Mobile Manipulation: a mobile robot manipulator with an RGB-D camera
• Active Semantic Mapping: semantic mapping of horticultural crops with a mobile manipulator (Cuaran, Ahluwalia, Koe, Uppalapati, Chowdhary, ICRA 2025)
• Viewpoint Planning: candidate viewpoints scored with an information utility that accounts for semantics and occlusions
• Efficient Ray Casting: for computing the information gain of each viewpoint

Course Relevance:

• Assignment 4: Mobile manipulator setup and control
• Module 7: SLAM for greenhouse navigation
• Module 8: Semantic mapping and scene understanding
• Extra Credit: Advanced perception with SAM & RTAB-Map

TA reference runs, not student submissions

Recorded TA runs of the course's Husky + UR3 mobile manipulator in Gazebo Harmonic (ROS 2 Jazzy, NVIDIA DGX Spark, July 2026): the Part 1 setup with a scripted base drive, and the Part 2 SLAM-on-Mars run (scripted drive, EKF sensor fusion and SLAM Toolbox), ending with a saved occupancy map. October 2026 re-runs on an MSI x86 laptop (Part 2: 15 of 15 waypoints reached, ground-truth error mean 0.256 m, max 0.395 m) and an RTAB-Map extra-credit run on the DGX Spark (trajectory error 0.271 m RMSE vs 0.725 m for wheel odometry) are on the SLAM-ing Mars page.

Watch the TA reference runs →

Final SLAM Toolbox occupancy map from the Part 2 TA reference run.