VisioBot
VisioBot explores the link between perception and motion: how a robot follows a route, detects a target, hands over control, and returns to its navigation task.
- Patrol
- Detect
- Approach
Implementation overview · See the source links below
The problem
Connect waypoint navigation with visual tracking without treating perception and movement as unrelated demonstrations.
Inside the implementation
The ROS 2 workspace includes a robot description, Gazebo worlds, Nav2 configuration and vision nodes. The capstone controller moves between patrol, visual approach and cooldown states.
Separate responsibilities
Robot configuration, navigation and vision live in focused ROS 2 packages. This makes the interactions between subsystems explicit.
An explicit handoff
The capstone cancels the navigation task when a target is acquired and uses visual feedback for approach behavior.
Simulation as a test bed
Gazebo and ROS tooling provide an environment for iterating on the behavior before making claims about a physical deployment.
What the evidence shows
The linked repository contains the implementation and supporting artifacts for this project. Follow the source links to inspect the details.
Current limitations
This case study describes the repository implementation, not verified physical autonomy. Bounding-box height is used as a distance proxy in the sampled controller, and measured navigation results are not yet documented here.
The next iteration
Record a reproducible simulation run, including target loss and recovery, and evaluate a more reliable distance estimate.
Let’s build something.
Interested in robotics, an idea for a useful tool, or something we could build together? I'd like to hear about it.