Autonomous systems
PathFinder — Autonomous Rover
A rover that learns to navigate complex terrain — trained in simulation, deployed on hardware.
PathFinder is GENBOT's autonomous rover project. It is designed to navigate complex terrain using a combination of ultrasonic sensing and reinforcement-learning policies that are trained in simulation before being deployed to the real robot.
Why simulation-first
Training a navigation policy directly on hardware is slow and risky. The PathFinder team trains policies in simulation where the rover can fail safely and iterate quickly, then deploys the learned policy onto the physical rover and evaluates it against real terrain.
Sensing and control
Ultrasonic sensors give PathFinder short-range obstacle awareness for navigation and edge detection. The learned policy maps these sensor readings to drive commands, letting the rover pick paths through terrain it has not seen during training.
Technologies
- Reinforcement learning for navigation policy training.
- Ultrasonic sensing for obstacle and edge detection.
- Embedded systems for motor control and sensor integration.
- Simulation environment for training before deployment.
Project status
PathFinder is an in-progress project. The team is iterating between simulation training and hardware trials, and every new sensor reading on the physical rover is fed back into the loop.
Related work
- NexBot — adaptive learning systems with real-time environment response.
- Smart Arena Bot — drivetrain and telemetry engineering from the same mechanical team.
Core technologies
Reinforcement learning · ultrasonic sensors · simulation-to-real transfer · embedded motor control
Status
In progress — simulation and hardware trials
Links
Want to train the next policy?
PathFinder's AI and embedded teams are open to new members every semester.
Join GENBOT