CrackRobotics

Crack robotics, one program at a time.

Write Python and watch the arm run it, in a MuJoCo physics simulation right here in your browser. You pass when the robot does the job, not when your code looks right.

Start Pick & Place, freeSee all lessons
main.pysim time 9.40 s
1from robot import arm, gripper, world, phase
2 
3ball = world.ball("red")
4cup = world.cup
5 
6phase("pre-grasp")
7arm.move_to(ball.position + [0, 0, 0.10])
8gripper.open()
9phase("approach")
10arm.move_to(ball.position)
11phase("grasp")
12gripper.close()
13phase("lift")
14arm.move_to(ball.position + [0, 0, 0.20])
15phase("transit")
16arm.move_to(cup.position + [0, 0, 0.20])
17phase("place")
18arm.move_to(cup.position + [0, 0, cup.height + 0.05])
19phase("release")
20gripper.open()
21phase("retreat")
22arm.move_to(cup.position + [0, 0, 0.20])
  • Grasped the red ball
  • Gripper released the ball
  • The red ball is resting in the cup

Step 4 of Pick & Place, replayed. You write this program yourself, one phase at a time.

Start with Pick & Place

Free in full, about an hour, and no robotics background needed.

Lesson overview
  1. 1Joints, targets & PD controlCommand joint angles, let physics run, and meet the PD controller behind every industrial arm.
  2. 2Forward kinematicsPredict exactly where the gripper is from the joint angles, using nothing but trigonometry.
  3. 3Inverse kinematics & straight-line motionTurn positions into joint angles, then move the gripper along straight lines with smooth, jerk-free timing.
  4. 4Your first pick & placeChain the classic manipulation phases into a program that moves the red ball into the cup.
  5. 5Clear the tableGeneralise your program: perceive every ball, pick each one, verify, retry, and survive a new layout on every run.
  6. ★Bonus: build your own IKReplace the black box. Derive the analytic inverse kinematics with the law of cosines and run a full pick with it.

Then keep going

Each lesson uses the same arm and the same Python API, so what you learn in one carries into the next.

Rather practise one algorithm at a time? Try the 22 problems, from forward kinematics to policy gradients.

Pick & Place is free, with no account. Pro unlocks every other lesson for $39 a month.

Start Pick & Place, free