Beyond what
we know.

How much can we learn about a robot before putting it in the water—or on the road? CRANE sim is the simulation foundation of the Cross-Domain Robotics Autonomous Navigation Engine. Composable physics modules model vehicle dynamics, actuation, buoyancy, drag, wind, and currents across ground, surface, and underwater platforms. Configurable simulated sensors—including LiDAR, cameras, IMUs, and positioning systems—support testing navigation and localization under controlled conditions. Shared ROS interfaces connect simulated and physical hardware, allowing navigation components to be tested and baseline parameters tuned in Unity before transferring to a real vehicle. The project explores reusable autonomy across different vehicle configurations, with simulation-first validation intended to reduce deployment risk and repeated hardware iteration. This project advanced to the Texas Science & Engineering Fair (TXSEF).

What if a robot could explore more than one route at once? Inspired by Stanford research on soft, tip-growing vine robots, this project develops an autonomous branching system for navigating and mapping complex, confined environments. Independently controlled branches can pursue separate paths simultaneously, reducing the need for sequential exploration and retraction. The design combines regulated pneumatic growth, cable-driven steering, depth-camera and IMU feedback, and a ROS 2-based controller. Parametric mechanical components and modular electronics support multiple branches, while experiments examine growth speed, branching response, energy use, and exploration time. The research investigates whether parallel branching can dramatically reduce navigation time as environments become more complex. The benefit depends on the number of available branches, path structure, and branching and retraction overhead—not a universal exponential speedup. This project advanced to the Texas Science & Engineering Fair (TXSEF).

How can simulation make autonomous navigation more reliable in the real world? Building on CRANE sim, this ongoing research tests ROS 2/Nav2 navigation against sensor errors, obstacles, and changing environmental conditions. Automated tuning explores safer, more efficient configurations while keeping classical planning and control at the core. Experiment replays and parameter visualizations will help explain where navigation fails, which settings matter, and how performance trades off against safety. The goal is a navigation stack that is better understood and better prepared for real-world deployment.