Skip to main content

Simulation · navigation sandbox

Status: Simulation

Operate a simulated robot.

A small differential-drive robot on a 16 × 10 occupancy grid. Choose a task, pick a target, and watch it plan a collision-free route with A* and follow it — turning in place, then driving.

Simulation mode — no physical robot connected

Loading the simulation…

Click or tap a free cell to set a navigation target. With the grid focused, use the arrow keys to move the cursor and press Enter to set it.

  • Robot
  • Planned path
  • Travelled
  • Target
  • Obstacle
Task
0.75 m/s

Task status

Idle

Target 14,1

Position X
0.50 m
Position Y
0.50 m
Heading
0°
Distance
0.0 m
Sim time
0.0 s
Activity
Standby

Readouts come from the simulation itself — idealised, no sensors.

Standing by at the start cell.

Everything here runs locally in your browser. It controls no hardware, uses an idealised model with perfect localisation and no sensor noise, and is not evidence of real-world robot autonomy.

01What it demonstrates

Small, real, and tested.

  • Path planning around fixed obstacles on an occupancy grid
  • A simple waypoint follower: turn toward the next cell, then drive
  • Replanning when the target changes mid-run
  • Explicit task states: idle, running, paused, complete, blocked
The Navigator two-wheel robot concept detecting an obstacle, stopping, turning and driving round it to a goal
Mechanism study · real-time 3DThe same machine as the console above: a differential-drive robot. Here the 3D concept scans, accelerates, detects a block, stops, re-plans and drives round it — its wheels turning exactly as its two-wheel kinematics require (no slip). The console's grid planner and this body describe one robot class.

Task state machine

Idle
Waiting for a task.
Running
The engine is advancing.
Paused
Stopped mid-task; Start resumes exactly where it left off.
Completed
A finite task (navigate, inspect) has finished.
Blocked
The planner could not find a route to the goal.

02Validation workflow

From browser to hardware — only when justified.

The sandbox is step one. Each later step needs results from the one before it, and hardware only ever runs under direct human supervision.

Stage 01
Start00:00
Stage 02
Route00:01.4
Stage 03
Target00:04.3
Source
1280 × 72030fps · 0.20MB

Still frame

Concept visualizationA simulated environment pictured as a light table: obstacles, a start point and a planned route. A rendering of the idea — not Moonpie's simulator and not real data.
  1. Step 1 · you can try this

    Software experiments

    Status: Simulation

    Planners and models evaluated against recorded and synthetic data. The browser sandbox is the first of these.

  2. Step 2

    Simulated environments

    Status: Proposed

    The same logic run in physics simulation (candidates: ROS 2, Gazebo), with failures logged and studied.

  3. Step 3

    Supervised hardware prototypes

    Status: Concept

    Only after simulation results justify it, and only under direct human supervision with stop controls.

03Method

Repeatable by design.

  • Deterministic and tested

    The same inputs give the same run, so a failure can be reproduced and studied. The sandbox engine ships with unit tests.

  • Human in the loop

    Autonomy is supervised. A person can pause, stop or override, and plans are approved before they are executed.