These are simplified, teaching-scale simulations — not the real ROS2/Nav2 stack — built to give you fast, visual intuition for ideas that are usually only explained in equations. Drag things, click things, break things.
Set independent left/right wheel speeds and watch how the robot's linear velocity (v) and angular velocity (ω) fall out of the difference between them — the same math behind /odom.
Equal speeds → straight line. Opposite speeds → rotation in place. Different-magnitude same-sign speeds → an arc turn.
The controller is trying to drive a simulated motor to a target velocity (dashed line). Push the gains too far and you'll see overshoot and oscillation — exactly what you'll fight when tuning a real drivetrain.
Higher Kd damps oscillation but reacts to noise. Higher Ki removes steady-state error but can overshoot.
This is what a global planner does on a costmap: search the known grid for the shortest collision-free path from start (S) to goal (G). The violet cells are nodes the search explored; the teal cells are the final path.
Click cells to add walls, then Run A*.
Try walling off the goal completely — the planner will correctly report no path found.
The global planner gives a route across the map — but the local planner has to dodge things in real time. This robot casts simulated range-sensor rays and steers using a simple attractive/repulsive blend, similar in spirit to DWA.
Click the canvas to add/remove obstacles, then Run.
Click an existing obstacle to remove it. Up to 8 obstacles at once.
Drag the orange prediction box (drag the body to move, the bottom-right corner to resize) and watch Intersection-over-Union update against the green ground-truth box. Adjust the threshold to see when a detection counts as a match.
IoU = overlap area ÷ combined area. A typical mAP@0.5 benchmark counts anything ≥ 0.5 as a true positive.
Every autonomous move a ROS2 robot makes flows through this same chain of nodes and topics. Click Step (or Auto-play) to walk through one full sense → plan → act cycle.
Six questions pulled straight from the labs above and the core curriculum.
Everything simulated above runs for real on the DIY build — under ₹50,000 in parts.