arm_push_easy.py

import jax
from brax import base
from jax import numpy as jnp

from envs.manipulation.arm_envs import ArmEnvs

"""
Push-Easy: Move a cube from a random location on the blue region to a random goal on the adjacent red region. The regions are very small.
- Observation space: 18-dim obs + 3-dim goal.
- Action space:      5-dim, each element in [-1, 1], corresponding to target angles for joints 1, 2, 4, 6, and finger closedness.

See _get_obs() and ArmEnvs._convert_action() for details.
"""


class ArmPushEasy(ArmEnvs):
    def _get_xml_path(self):
        return "envs/assets/panda_push_easy.xml"

    @property
    def action_size(self) -> int:
        return 5  # Override default (actuator count)

    # See ArmEnvs._set_environment_attributes for descriptions of attributes
    def _set_environment_attributes(self):
        self.env_name = "arm_push_easy"
        self.episode_length = 150

        self.goal_indices = jnp.array([0, 1, 2])  # Cube position
        self.completion_goal_indices = jnp.array([0, 1, 2])  # Identical
        self.state_dim = 18
        self.goal_reach_thresh = 0.1

        self.arm_noise_scale = 0
        self.cube_noise_scale = 0.1
        self.goal_noise_scale = 0.1

    def _get_initial_state(self, rng):
        rng, subkey1, subkey2 = jax.random.split(rng, 3)
        cube_q_xy = self.sys.init_q[:2] + self.cube_noise_scale * jax.random.uniform(
            subkey1, [2], minval=-1
        )
        cube_q_remaining = self.sys.init_q[2:7]
        target_q = self.sys.init_q[7:14]
        arm_q_default = jnp.array(
            [1.571, 0.742, 0, -1.571, 0, 3.054, 1.449, 0.04, 0.04]
        )  # Start closer to the relevant area
        arm_q = arm_q_default + self.arm_noise_scale * jax.random.uniform(
            subkey2, [self.sys.q_size() - 14], minval=-1
        )

        q = jnp.concatenate([cube_q_xy] + [cube_q_remaining] + [target_q] + [arm_q])
        qd = jnp.zeros([self.sys.qd_size()])
        return q, qd

    def _get_initial_goal(self, pipeline_state: base.State, rng):
        rng, subkey = jax.random.split(rng)
        cube_goal_pos = jnp.array([0.1, 0.6, 0.03]) + jnp.array(
            [self.goal_noise_scale, self.goal_noise_scale, 0]
        ) * jax.random.uniform(subkey, [3], minval=-1)
        return cube_goal_pos

    def _compute_goal_completion(self, obs, goal):
        # Goal occupancy: is the cube close enough to the goal?
        current_cube_pos = obs[self.completion_goal_indices]
        goal_pos = goal[:3]
        dist = jnp.linalg.norm(current_cube_pos - goal_pos)

        success = jnp.array(dist < self.goal_reach_thresh, dtype=float)
        success_easy = jnp.array(dist < 0.3, dtype=float)
        success_hard = jnp.array(dist < 0.03, dtype=float)
        return success, success_easy, success_hard

    def _update_goal_visualization(
        self, pipeline_state: base.State, goal: jax.Array
    ) -> base.State:
        updated_q = pipeline_state.q.at[7:10].set(
            goal[:3]
        )  # Only set the position, not orientation
        updated_pipeline_state = pipeline_state.replace(qpos=updated_q)
        return updated_pipeline_state

    def _get_obs(
        self, pipeline_state: base.State, goal: jax.Array, timestep
    ) -> jax.Array:
        """
        Observation space (18-dim)
         - q_subset (10-dim): 3-dim cube position, 7-dim joint angles
         - End-effector (6-dim): position and velocity
         - Fingers (2-dim): finger distance, gripper force
        Note q is 23-dim: 7-dim cube position/angle, 7-dim goal marker position/angle, 7-dim joint angles, 2-dim finger offset

        Goal space (3-dim): position of cube
        """
        q_indices = jnp.array([0, 1, 2, 14, 15, 16, 17, 18, 19, 20])
        q_subset = pipeline_state.q[q_indices]

        eef_index = 8  # Cube is 0, goal marker is 1, then links 1-7 are indices 2-8. The end-effector (eef) base is merged with link 7, so we say link 7 index = eef index.
        eef_x_pos = pipeline_state.x.pos[eef_index]
        eef_xd_vel = pipeline_state.xd.vel[eef_index]

        left_finger_index = 9
        left_finger_x_pos = pipeline_state.x.pos[left_finger_index]
        right_finger_index = 10
        right_finger_x_pos = pipeline_state.x.pos[right_finger_index]
        finger_distance = jnp.linalg.norm(right_finger_x_pos - left_finger_x_pos)[
            None
        ]  # [None] expands dims from 0 to 1
        gripper_force = (pipeline_state.qfrc_actuator[:-2]).mean(
            keepdims=True
        ) * 0.1  # Normalize it from range [-20, 20] to [-2, 2]

        return jnp.concatenate(
            [q_subset]
            + [eef_x_pos]
            + [eef_xd_vel]
            + [finger_distance]
            + [gripper_force]
            + [goal]
        )

    def _get_arm_angles(self, pipeline_state: base.State) -> jax.Array:
        q_indices = jnp.arange(14, 21)
        return pipeline_state.q[q_indices]