Push an object that cannot be grasped into a marked zone: Unitree Go2 in the workshop bench
push-to-zone--unitree-go2--workshop-bench--s4remixUnitree Go2Manipulationmedium
Instruction
You control a Unitree Go2 (standing on the floor at (0.3, -1.4), facing 90 degrees). A 15 kg quadruped driven by a model-based trot controller (body velocity commands); a forward head camera; it can push things with its body. The scene is a workshop: a 0.9 m workbench under a pegboard, a parts cart, a tool cabinet and a drill press. Arms mount at the back of the bench. Everything is simulated with MuJoCo physics: objects are free rigid bodies that move only through contact and friction, so a loose grip or a collision can drop or knock things over. World frame in metres: +x east, +y north, +z up, floor at z = 0 (the floor area spans x -2.6..2.6, y -2.2..1.95); headings in degrees counter-clockwise from +x.
Task
The carton is too wide to grasp. Push it into the green zone.
Success: the centre of the carton is inside the green zone, it is upright (tilted less than 15 degrees) and nothing of the robot touches it; and no safety event happened (a hard collision with furniture or walls, an object falling to the floor, the robot falling over, or harm to a person; safety_events in robo observe lists them). Judged by the episode server from the simulated state after you call robo done and the robot has held still for 10 steps.
Controls. robo act VX VY WZ (3 values; one step is 50 ms; --repeat N holds it for N steps). base.twist VX, VY, WZ ([-0.6, -0.4, -69] to [0.6, 0.4, 69]; m/s, m/s, deg/s): body-frame velocity command for the trot controller (x forward, y left, WZ counter-clockwise); all zero = stop stepping and stand. All zero holds still. Skills (robo skill NAME ARG ...; robo info lists them with their arguments): go_to X Y [YAW_DEG=none] [SPEED=0.5]: walk to (X, Y): turn towards it, trot along the straight line, stop (then turn to YAW_DEG); it does not plan around obstacles (give it waypoints) and reports if blocked; turn YAW_DEG: turn in place to heading YAW_DEG (0 = +x east, 90 = +y north); look_at X Y Z: turn to face (X, Y) so the head camera looks at it, and hold still for a second.
Observation. robo observe reports robot (what the robot is, its capabilities and grasp, its workspace, and its current state: position, heading and motion state), objects (per task object: kind, colour, centre pos, tilt, yaw, size, mass and resting_on: a surface, a fixture, another object, gripper or the floor), fixtures (containers with their inner size and rim height, pegs, marks, pads, tags), scene (surfaces with their top heights and sizes, obstacles, the floor area) and safety_events. Cameras (robo observe --image --camera NAME): overview, chase, robot/head, bench_cam, top.
The step budget is 2400 steps (120 s of simulated time).
How the robot is controlled and scored
You are controlling a simulated robot. Read the task below, then solve it by running the robo command in your shell (start with robo info and robo observe). Keep going until the task is done, then call robo done once. Do not stop to ask questions; there is no human to answer.
How to control the robot
You are the robot's policy. You act only through the robo command in your shell. There is no other way to move the robot, and you cannot read or change the simulator, the scoring, or other files to succeed; the episode server judges the final physical state itself.
robo info # the robot, its sensors, action groups, skills and step budget
robo observe # robot and scene state as numbers
robo observe --image [--camera C] # also saves a camera image and prints its path (open it to look)
robo act V1 V2 ... [--repeat N] # one low-level action (the action groups under Controls), applied N times (N <= 50)
robo skill NAME ARG ... # run a skill listed by `robo info`; it runs until it finishes and reports the result
robo done "short summary" # end the episode and ask for scoring
robo give-up "reason" # end the episode without claiming success- Positions are in metres in the world frame (+z up); angles are in degrees unless a field says otherwise.
- The episode has a fixed step budget (see
robo info); every simulated control step counts, including the steps a skill runs. - Skills are ordinary controllers: they can fail, stop early or be blocked by the scene. Read what they report and re-observe.
- Success is judged about 10 steps after you call
robo done, with the robot holding still (each action group's hold value: zero for velocity and delta commands, full brake for a car), so the goal must still be true when the robot stops. - Call
robo doneexactly once when finished.
Run this task
bench eval run \
-d benchflow/robouse-composed@0.1 \
--registry https://robouse.ai/hub/registry.json \
--agent oracle \
--include push-to-zone--unitree-go2--workshop-bench--s4Pinned to robohub commit 1f901b956f61. The verifier and the reference solution are not published.