Camden Boston Engineering

Independent browser simulation · 2026

Autonomous Navigation Lab

Choose a destination, change the obstacles, and inspect how a mobile robot plans a route and responds to moving objects.

2022 → 2026

Original ROS & Gazebo project
Rebuilt as an interactive browser model

A known room, three moving obstacles, and a destination you choose.

Try this: Run the tour, increase obstacle motion, then switch controllers and reset. Watch the gap and contact count.

Live 2D model · ideal robot pose and obstacle tracking · simulation time

The engineering

See the decision,
not just the motion.

The original project combined autonomous mapping, A* navigation, and moving-obstacle response in ROS. This rebuild makes each step visible and separates the model from the interface.

01

Observe & map

Lidar rays mark occupied and free cells. In Map mode, the robot routes toward reachable frontiers—the edge of what it has observed. Unknown space stays shaded.

02

Plan with room to spare

A* or Dijkstra searches an inflated occupancy grid, keeps diagonal moves out of corners, and simplifies the route with clearance checks. The robot stops by actual distance to its goal.

03

Anticipate the crossing

The predictive controller tests forward, turning, waiting, and escape motions against constant-velocity obstacle predictions. It executes a short step, observes again, and updates the plan.

Measured, reproducible results

More clearance.
A deliberate tradeoff.

Both controllers use the same new route planner, ideal pose, room, goal sequence, and collision checks. The reactive comparison isolates a narrow front-distance rule inspired by the historical implementation; it is not a replay of the original ROS controller.

Fixed six-goal tours · 21 obstacle conditions
Measurement Predictive Reactive
Benchmark Loading checked results…

Seven starting phases × three obstacle speeds. Robot speed limit 0.8 m/s.

Full benchmark data ↓

What changed since 2022

The recovered source mixed map conversion, search, control, and ROS transport. This version separates geometry, sensing, planning, control, simulation, and display. It adds footprint-aware routes, precise goal stopping, predictive avoidance, automated tests, and exportable run data.

What this demonstration represents

A new browser simulation based on the original room and obstacle motions. It uses ideal pose and tracked obstacle velocity, kinematic motion, and geometric contact checks. Mapping uses known pose; this browser does not run ROS, GMapping, AMCL, or full SLAM. Results establish behavior in this model, not real-robot performance or a measured improvement over the old videos.