A* (A-Star): finds the shortest path using a heuristic to guide search. Combines Dijkstra's cost-so-far with an estimated distance-to-goal (Manhattan/Euclidean).

RRT (Rapidly-Exploring Random Trees): randomly grows a tree from start, biased toward the goal. Ideal for high-dimensional configuration spaces.

Click canvas to place obstacles. Press Run to watch the algorithm find a path.

Robots "see" the world through a sequence of image processing steps:

1. Grayscale → reduce colour channels
2. Gaussian blur → reduce noise
3. Edge detection → Sobel / Canny operators
4. Contour / object detection → bounding boxes
5. Depth estimation → stereo or LiDAR-style point cloud

Drag objects on the canvas to see real-time detection.

PID Controller: the backbone of robot motion control.
P (Proportional) — react to current error
I (Integral) — correct accumulated error
D (Derivative) — dampen oscillation

Reinforcement Learning: an agent learns to balance a cart-pole by receiving rewards for staying upright. The neural network maps state → action.

Drag the setpoint to challenge the controller.

Swarm robots follow simple local rules that produce complex emergent behaviour:

Separation: avoid crowding neighbours
Alignment: steer toward average heading
Cohesion: move toward group centre

This is Boids flocking — also used to model fish schools, bird murmurations, and drone swarms. Each robot has no global knowledge, yet the group self-organises.

Click canvas to place obstacles / food targets.