RX200 — Push

The arm must push a 4 cm cube across the cafe-table to a goal point on the table top. Closed gripper acts as a flat paddle.

Env IDs: RX200PushSim-v0 / RX200PushGoalSim-v0 / RX200PushReal-v0 / RX200PushGoalReal-v0. Sim-only ZED 2 variants: RX200Zed2PushSim-v0 / RX200Zed2PushGoalSim-v0.

Description

RX200 flush-mounted on the cafe-table. At reset the arm goes to home pose, the gripper closes (init_close_gripper = [0.018, -0.018] m), and a red cube spawns on the cafe-table top.

Action Space

Joint mode (default). Box(5,) — same 5-joint command as RX200 — Reach (waist / shoulder / elbow / wrist_angle / wrist_rotate). Gripper is NOT in the action vector for push.

EE mode (ee_action_type=True). Box(3,) — Δ EE position.

Observation Space

Standard env. Box. Extends the reach obs with cube state:

  • EE position (3, m)

  • EE rpy (3, rad)

  • Unit vector cube → goal (3, normalised)

  • Distance cube → goal (1, m)

  • Current joint positions (8 — alphabetical: elbow, gripper, left_finger, right_finger, shoulder, waist, wrist_angle, wrist_rotate)

  • Previous action (5 or 3)

  • Current joint velocities (8)

  • Cube position in base frame (3, m)

  • Cube rpy (3, rad)

  • Cube linear velocity (3, m/s, finite-difference)

  • Cube angular velocity (3, rad/s, finite-difference)

  • Cube position relative to EE (3, m)

Goal env. Dict. desired_goal = Box(3,) on table top (x ∈ [0.15, 0.35], y ∈ [-0.15, 0.15], z ≈ 0.015). achieved_goal = Box(3,) cube XYZ.

Rewards

Sparse: 0.0 if ‖cube − goal‖ < reach_tolerance else -1.0.

Dense: same shape as RX200 — Reach, but distance is from CUBE to goal (not EE to goal). Defaults from config/rx200_push_task_config.yaml.

Starting State

Joint pose: zeros (URDF home). Gripper closed at init_close_gripper = [0.018, -0.018] m.

Cube spawn. 4 cm red cube at default cube_init_pos = [0.25, 0.0, 0.015] in base frame. Randomised XY if random_cube_spawn=True.

Goal sampling. Push goal ∈ Box(3,) on the table top within position_goal_min/max.

Episode End

Truncation. max_episode_steps (default 100). Real env on stale /rx200/joint_states.

Termination. ‖cube − goal‖ < reach_tolerance (sparse only).

Arguments

Inherits RX200 — Reach kwargs plus push-specific random_cube_spawn / random_goal / (real only) cube_pose_topic (default /cube_pose), cube_pose_timeout_s (1.0 s), auto_launch_cube_tracker / cube_tracker_camera / cube_tracker_target_frame.

Version History

  • v0 — first release (rl_environments v0.1.0).