NED2 — Pick-and-Place

The arm must grasp a 4 cm cube, lift it, and place it at a target point. Action vector gains one extra scalar at action[-1] controlling the mors gripper. Optional multi_goal lift-then-place curriculum.

Env IDs: NED2PnPSim-v0 / NED2PnPGoalSim-v0 / NED2PnPReal-v0 / NED2PnPGoalReal-v0.

Description

Niryo Ned2 6-DoF arm flush on the cafe-table. At reset the arm goes to init pose, the gripper opens, and the cube spawns on the table. is_grasped triggers when ‖cube − ee‖ < grasp_dist_thresh AND mors_1_pos < grasp_finger_thresh.

Action Space

Joint mode (default). Box(7,):

  • Indices 0–5: 6-joint arm command (same as NED2 — Reach).

  • Index 6: gripper command (absolute mors_1 position) ∈ [-0.01, +0.01] m. On sim, the trajectory is published directly to gazebo_tool_commander with both mors mirrored. On real, the scalar is thresholded to a binary niryo_robot_tools_commander “open”/”close” action.

EE mode (ee_action_type=True). Box(4,) — Δ EE position + gripper command.

Observation Space

Standard env. Box. Extends push obs with is_grasped (1 dim, 0/1 float).

Goal env. Dict. desired_goal = Box(3,) with z up to lift height. achieved_goal = cube XYZ.

Rewards

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

Dense with multi_goal=True and grasp shaping:

reward = -multiplier_dist_reward * ‖cube − current_goal‖
       + grasp_bonus if is_grasped
       + reached_goal_reward if ‖cube − pnp_goal‖ < reach_tolerance
       + step_reward + (joint/none/goal-space penalties)

Defaults from config/ned2_pnp_task_config.yaml: grasp_dist_thresh=0.05, grasp_finger_thresh = 0.0 (m — placeholder; calibrate empirically before relying on ``is_grasped``), lift_height=0.15.

env = gym.make("NED2PnPSim-v0", reward_type="Dense", multi_goal=True)
env = gym.make("NED2PnPGoalSim-v0", reward_type="Sparse", multi_goal=True)

Starting State

Joint pose: URDF zero. Gripper open (init_open_gripper = [0.01, 0.01] m).

Cube spawn. 4 cm red cube at default cube_init_pos = [0.25, 0.0, 0.015] in base frame. Note: sim env has [0.180, 0.000, 0.015] hardcoded fallback — inconsistent with YAML; align before relying on the static spawn.

Goal sampling. pnp_goal ∈ Box(3,) from position_goal_min/max. With multi_goal=True, intermediate_goal = cube_init + [0, 0, 0.15].

Episode End

Truncation. max_episode_steps (default 100). Real env on stale /ned2/joint_states.

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

Arguments

Inherits NED2 — Reach and NED2 — Push kwargs plus pnp-specific:

  • multi_goal (bool, default True) — lift-then-place curriculum.

  • lift_height (float, default 0.15 m) — vertical offset of intermediate_goal above cube spawn.

  • use_wrist_camera (bool, default False) — opt-in Niryo wrist camera subscriber for close-up perception.

Version History

  • v0 — first release (rl_environments v0.1.0). grasp_finger_thresh = 0.0 is a placeholder — measure the closed mors position on the actual hardware and update the YAML before relying on is_grasped in real training.