Small decisions. Interesting possibilities.Submit contentSubmit

Jev robot control

Recorded xArm7 MuJoCo pick-and-place trial comparing Jev with two language-model controllers under the same Cartesian action interface.

Added to Jevfast

How it uses Jev

Each robot-control cycle asks Jev to choose a pick-and-place intent from measured geometry and contacts, then asks a second Jev request for X/Y/Z movement signs and the gripper command. A deterministic executor applies one bounded motion increment.

What Jev decides

Intent selectionExample answers · not a recorded Jev response · Source ↗
Question 1 · intent
YOUR APP
INSTRUCTION

Choose the next immediate pick-and-place intent from actual geometry and contact feedback.

STATE

Structured MuJoCo geometry, signed target-minus-tool errors, alignment and bilateral contact checks, and current action context; not camera pixels.

JEV · CHOICE
  1. approach
  2. grasp
  3. lift
  4. carry
  5. lower
  6. release
  7. withdraw
  8. finish
Motor incrementExample answers · not a recorded Jev response · Source ↗
Question 1 · x
YOUR APP
INSTRUCTION

Choose X-axis motion for the selected intent.

STATE

The same measured state plus selected_intent, per-axis current position and target errors.

JEV · CHOICE
  1. negative
  2. hold
  3. positive
Question 2 · y
YOUR APP
INSTRUCTION

Choose Y-axis motion for the selected intent.

STATE

The same measured state plus selected_intent, per-axis current position and target errors.

JEV · CHOICE
  1. negative
  2. hold
  3. positive
Question 3 · z
YOUR APP
INSTRUCTION

Choose Z-axis motion for the selected intent.

STATE

The same measured state plus selected_intent, per-axis current position and target errors.

JEV · CHOICE
  1. negative
  2. hold
  3. positive
Question 4 · gripper
YOUR APP
INSTRUCTION

Choose the gripper command for the selected intent.

STATE

The same measured state plus selected_intent, per-axis current position and target errors.

JEV · CHOICE
  1. open
  2. hold
  3. close

App workflow

  1. Observe the scene

    Read structured simulator geometry and contact feedback for the current robot state.

  2. Choose intent and motor channels

    Ask Jev for one intent Choice and then a second request for axis and gripper Choices.

  3. Execute and verify

    Apply one bounded Cartesian increment, observe the next state, and record responses and trajectories for replay.

Why it is interesting

The same two-stage Cartesian interface was recorded for Jev and comparison language models; the repository provides replayable answers and physical trajectories instead of only a promotional speed claim.