Swap two drones

crazyflie-swarm-swapcrazyflieBitcraze Crazyflie 2 quadrotorManipulationhard

Reference solution, 470 of 2,000 steps.

Instruction

You fly three Bitcraze Crazyflie 2 nano-quadrotors (cf0, cf1, cf2) (MuJoCo Menagerie model; 27 g and 9 cm across each) in an indoor flight arena. World frame in metres: +x east, +y north, +z up, floor at z = 0. The arena is 8 m x 6 m (x from -4 to 4, y from -3 to 3) and 3 m high, closed by walls and a ceiling. A full-height partition wall at x = 1.5 separates the west hall from the east room; the only way through it is a 0.36 m x 0.36 m window centred at (1.5, 1.8, 1.0). The west hall has a home pad, a teal pad A, three square racing gates (orange frames, 0.5 m openings) and three swarm pads S0, S1, S2; the east room has a yellow pad B and a 0.9 m high shelf with a white landing platform (0.3 m square) on top. Pads are 0.3 m squares. Every position is listed under arena in robo observe. Touching anything other than the floor, a pad or the platform is a crash and fails the task; so does a collision between drones.

Task

Three drones rest on pads: cf0 on S0, cf1 on S1, cf2 on S2. Swap cf0 and cf2: cf0 must end on S2 and cf2 on S0, both landed; cf1 must end on S1 (it may move in between).

Success: cf0 rests on S2, cf1 on S1 and cf2 on S0; any crash, or two drones ever closer than 0.2 m (centre to centre), fails the task. Judged by the episode server from the simulated state after you call robo done and the drones have held still for 10 steps. A drone "rests on" a pad or the platform when it touches it, moves slower than 0.1 m/s and is tilted less than 20 degrees.

Controls. robo act VX0 VY0 VZ0 VX1 VY1 VZ1 VX2 VY2 VZ2 sets the velocity setpoints of drones cf0, cf1 and cf2 for the next step(s), each in [-1, 1] (x 1 m/s along world x, y, z); headings stay fixed. One step is 50 ms; --repeat N holds the setpoints for N steps. A drone whose three values are all zero holds its position; resting on a surface with VZ <= 0 its motors stay off. Skills: takeoff_all [Z], goto ID X Y Z [TOL] (one drone flies in a straight line while the others hold position), goto_all X0 Y0 Z0 X1 Y1 Z1 X2 Y2 Z2 [TOL] (all at once, straight lines, no collision avoidance), land ID, land_all and hover [SECONDS].

Observation. robo observe reports drones (per drone: id, pos, vel, yaw_deg, roll_deg, pitch_deg, landed_on, motors_on), arena and safety_events. Cameras: swarm (default, the swarm pads from the south), cf0/chase, cf1/chase, cf2/chase, overview (the west hall) and east_room.

The step budget is 2000 steps (100 s of flight).

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 done exactly once when finished.

Run this task

ROBOUSE_ORACLE_TOKEN=$(openssl rand -hex 16) \
  bench eval run \
    -d benchflow/crazyflie@0.2 \
    --registry https://robouse.ai/hub/registry.json \
    --agent oracle \
    --include crazyflie-swarm-swap

Embodiment

Bitcraze Crazyflie 2 quadrotor

Drone · MuJoCo Menagerie models

Action · 9 numbers per step

VX0VX01VY0VY01VZ0VZ01VX1VX11VY1VY11VZ1VZ11VX2VX21VY2VY21VZ2VZ21

cf0.velocity_setpoint (velocity_setpoint, x 1 m/s (world)): zero = hold position | cf1.velocity_setpoint (velocity_setpoint, x 1 m/s (world)): zero = hold position | cf2.velocity_setpoint (velocity_setpoint, x 1 m/s (world)): zero = hold position. One step = 50 ms.

Skills

takeoff-all [Z=1.0]: all drones climb straight up to height Z (at most 200 steps per call)goto ID X Y Z [TOL=0.05]: fly drone ID in a straight line to (X, Y, Z) while the others hold position (at most 300 steps per call)goto-all X0 Y0 Z0 X1 Y1 Z1 X2 Y2 Z2 [TOL=0.05]: fly all drones at once, each in a straight line to its point (X0 Y0 Z0 X1 Y1 Z1 X2 Y2 Z2); no collision avoidance (at most 300 steps per call)land ID: drone ID descends straight down and switches its motors off (at most 200 steps per call)land-all: all drones descend straight down and switch their motors off (at most 200 steps per call)hover [SECONDS=1.0]: all drones hold position (at most 100 steps per call)

Observation

State: positions of the robot and objects

dronestime_sarenasafety_events

Camera

swarm

Budget and scoring

2,000 stepsjudged on the final state

In datasets

benchflow/crazyflie

Pinned to robohub commit e472b1a1e041. The verifier and the reference solution are not published.

No trials yet.