feat(utils): precise_sleep() less CPU hungry without sacrificing accuracy (#2526)

docs/source/envhub_leisaac.mdx

@@ -139,7 +139,7 @@ from lerobot.teleoperators import (  # noqa: F401
     make_teleoperator_from_config,
     so101_leader,
 )
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import init_logging
 from lerobot.envs.factory import make_env
 
@@ -196,7 +196,7 @@ def teleop_loop(teleop: Teleoperator, env: gym.Env, fps: int):
             obs, info = env.reset()
 
         dt_s = time.perf_counter() - loop_start
-        busy_wait(1 / fps - dt_s)
+        precise_sleep(1 / fps - dt_s)
         loop_s = time.perf_counter() - loop_start
         print(f"\ntime: {loop_s * 1e3:.2f}ms ({1 / loop_s:.0f} Hz)")
 

docs/source/il_robots.mdx

@@ -393,7 +393,7 @@ import time
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.robots.so100_follower.config_so100_follower import SO100FollowerConfig
 from lerobot.robots.so100_follower.so100_follower import SO100Follower
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import log_say
 
 episode_idx = 0
@@ -415,7 +415,7 @@ for idx in range(dataset.num_frames):
     }
     robot.send_action(action)
 
-    busy_wait(1.0 / dataset.fps - (time.perf_counter() - t0))
+    precise_sleep(1.0 / dataset.fps - (time.perf_counter() - t0))
 
 robot.disconnect()
 ```

examples/backward_compatibility/replay.py

@@ -45,7 +45,7 @@ from lerobot.robots import (  # noqa: F401
     so101_follower,
 )
 from lerobot.utils.constants import ACTION
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import (
     init_logging,
     log_say,
@@ -97,7 +97,7 @@ def replay(cfg: ReplayConfig):
         robot.send_action(action)
 
         dt_s = time.perf_counter() - start_episode_t
-        busy_wait(1 / dataset.fps - dt_s)
+        precise_sleep(1 / dataset.fps - dt_s)
 
     robot.disconnect()
 

examples/lekiwi/replay.py

@@ -20,7 +20,7 @@ from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.robots.lekiwi.config_lekiwi import LeKiwiClientConfig
 from lerobot.robots.lekiwi.lekiwi_client import LeKiwiClient
 from lerobot.utils.constants import ACTION
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import log_say
 
 EPISODE_IDX = 0
@@ -58,7 +58,7 @@ def main():
         # Send action to robot
         _ = robot.send_action(action)
 
-        busy_wait(max(1.0 / dataset.fps - (time.perf_counter() - t0), 0.0))
+        precise_sleep(max(1.0 / dataset.fps - (time.perf_counter() - t0), 0.0))
 
     robot.disconnect()
 

examples/lekiwi/teleoperate.py

@@ -19,7 +19,7 @@ import time
 from lerobot.robots.lekiwi import LeKiwiClient, LeKiwiClientConfig
 from lerobot.teleoperators.keyboard.teleop_keyboard import KeyboardTeleop, KeyboardTeleopConfig
 from lerobot.teleoperators.so100_leader import SO100Leader, SO100LeaderConfig
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.visualization_utils import init_rerun, log_rerun_data
 
 FPS = 30
@@ -71,7 +71,7 @@ def main():
         # Visualize
         log_rerun_data(observation=observation, action=action)
 
-        busy_wait(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))
+        precise_sleep(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))
 
 
 if __name__ == "__main__":

examples/phone_to_so100/replay.py

@@ -29,7 +29,7 @@ from lerobot.robots.so100_follower.robot_kinematic_processor import (
 )
 from lerobot.robots.so100_follower.so100_follower import SO100Follower
 from lerobot.utils.constants import ACTION
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import log_say
 
 EPISODE_IDX = 0
@@ -96,7 +96,7 @@ def main():
         # Send action to robot
         _ = robot.send_action(joint_action)
 
-        busy_wait(1.0 / dataset.fps - (time.perf_counter() - t0))
+        precise_sleep(1.0 / dataset.fps - (time.perf_counter() - t0))
 
     # Clean up
     robot.disconnect()

examples/phone_to_so100/teleoperate.py

@@ -32,7 +32,7 @@ from lerobot.robots.so100_follower.so100_follower import SO100Follower
 from lerobot.teleoperators.phone.config_phone import PhoneConfig, PhoneOS
 from lerobot.teleoperators.phone.phone_processor import MapPhoneActionToRobotAction
 from lerobot.teleoperators.phone.teleop_phone import Phone
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.visualization_utils import init_rerun, log_rerun_data
 
 FPS = 30
@@ -114,7 +114,7 @@ def main():
         # Visualize
         log_rerun_data(observation=phone_obs, action=joint_action)
 
-        busy_wait(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))
+        precise_sleep(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))
 
 
 if __name__ == "__main__":

examples/so100_to_so100_EE/replay.py

@@ -30,7 +30,7 @@ from lerobot.robots.so100_follower.robot_kinematic_processor import (
 )
 from lerobot.robots.so100_follower.so100_follower import SO100Follower
 from lerobot.utils.constants import ACTION
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import log_say
 
 EPISODE_IDX = 0
@@ -97,7 +97,7 @@ def main():
         # Send action to robot
         _ = robot.send_action(joint_action)
 
-        busy_wait(1.0 / dataset.fps - (time.perf_counter() - t0))
+        precise_sleep(1.0 / dataset.fps - (time.perf_counter() - t0))
 
     # Clean up
     robot.disconnect()

examples/so100_to_so100_EE/teleoperate.py

@@ -32,7 +32,7 @@ from lerobot.robots.so100_follower.robot_kinematic_processor import (
 from lerobot.robots.so100_follower.so100_follower import SO100Follower
 from lerobot.teleoperators.so100_leader.config_so100_leader import SO100LeaderConfig
 from lerobot.teleoperators.so100_leader.so100_leader import SO100Leader
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.visualization_utils import init_rerun, log_rerun_data
 
 FPS = 30
@@ -120,7 +120,7 @@ def main():
         # Visualize
         log_rerun_data(observation=leader_ee_act, action=follower_joints_act)
 
-        busy_wait(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))
+        precise_sleep(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))
 
 
 if __name__ == "__main__":

src/lerobot/rl/actor.py

@@ -78,7 +78,7 @@ from lerobot.transport.utils import (
     transitions_to_bytes,
 )
 from lerobot.utils.random_utils import set_seed
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.transition import (
     Transition,
     move_state_dict_to_device,
@@ -398,7 +398,7 @@ def act_with_policy(
 
         if cfg.env.fps is not None:
             dt_time = time.perf_counter() - start_time
-            busy_wait(1 / cfg.env.fps - dt_time)
+            precise_sleep(1 / cfg.env.fps - dt_time)
 
 
 #  Communication Functions - Group all gRPC/messaging functions

src/lerobot/rl/gym_manipulator.py

@@ -74,7 +74,7 @@ from lerobot.teleoperators import (
 from lerobot.teleoperators.teleoperator import Teleoperator
 from lerobot.teleoperators.utils import TeleopEvents
 from lerobot.utils.constants import ACTION, DONE, OBS_IMAGES, OBS_STATE, REWARD
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import log_say
 
 logging.basicConfig(level=logging.INFO)
@@ -114,7 +114,7 @@ def reset_follower_position(robot_arm: Robot, target_position: np.ndarray) -> No
     for pose in trajectory:
         action_dict = dict(zip(current_position_dict, pose, strict=False))
         robot_arm.bus.sync_write("Goal_Position", action_dict)
-        busy_wait(0.015)
+        precise_sleep(0.015)
 
 
 class RobotEnv(gym.Env):
@@ -238,7 +238,7 @@ class RobotEnv(gym.Env):
             reset_follower_position(self.robot, np.array(self.reset_pose))
             log_say("Reset the environment done.", play_sounds=True)
 
-        busy_wait(self.reset_time_s - (time.perf_counter() - start_time))
+        precise_sleep(self.reset_time_s - (time.perf_counter() - start_time))
 
         super().reset(seed=seed, options=options)
 
@@ -713,7 +713,7 @@ def control_loop(
             transition = env_processor(transition)
 
         # Maintain fps timing
-        busy_wait(dt - (time.perf_counter() - step_start_time))
+        precise_sleep(dt - (time.perf_counter() - step_start_time))
 
     if dataset is not None and cfg.dataset.push_to_hub:
         logging.info("Pushing dataset to hub")
@@ -745,7 +745,7 @@ def replay_trajectory(
         )
         transition = action_processor(transition)
         env.step(transition[TransitionKey.ACTION])
-        busy_wait(1 / cfg.env.fps - (time.perf_counter() - start_time))
+        precise_sleep(1 / cfg.env.fps - (time.perf_counter() - start_time))
 
 
 @parser.wrap()

src/lerobot/scripts/lerobot_find_joint_limits.py

@@ -50,7 +50,7 @@ from lerobot.teleoperators import (  # noqa: F401
     make_teleoperator_from_config,
     so100_leader,
 )
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 
 
 @dataclass
@@ -114,7 +114,7 @@ def find_joint_and_ee_bounds(cfg: FindJointLimitsConfig):
             print(f"Min joint pos position {np.round(min_pos, 4).tolist()}")
             break
 
-        busy_wait(0.01)
+        precise_sleep(0.01)
 
 
 def main():

src/lerobot/scripts/lerobot_record.py

@@ -119,7 +119,7 @@ from lerobot.utils.control_utils import (
     sanity_check_dataset_robot_compatibility,
 )
 from lerobot.utils.import_utils import register_third_party_devices
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import (
     get_safe_torch_device,
     init_logging,
@@ -364,7 +364,7 @@ def record_loop(
             log_rerun_data(observation=obs_processed, action=action_values)
 
         dt_s = time.perf_counter() - start_loop_t
-        busy_wait(1 / fps - dt_s)
+        precise_sleep(1 / fps - dt_s)
 
         timestamp = time.perf_counter() - start_episode_t
 

src/lerobot/scripts/lerobot_replay.py

@@ -62,7 +62,7 @@ from lerobot.robots import (  # noqa: F401
 )
 from lerobot.utils.constants import ACTION
 from lerobot.utils.import_utils import register_third_party_devices
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import (
     init_logging,
     log_say,
@@ -121,7 +121,7 @@ def replay(cfg: ReplayConfig):
         _ = robot.send_action(processed_action)
 
         dt_s = time.perf_counter() - start_episode_t
-        busy_wait(1 / dataset.fps - dt_s)
+        precise_sleep(1 / dataset.fps - dt_s)
 
     robot.disconnect()
 

src/lerobot/scripts/lerobot_teleoperate.py

@@ -89,7 +89,7 @@ from lerobot.teleoperators import (  # noqa: F401
     so101_leader,
 )
 from lerobot.utils.import_utils import register_third_party_devices
-from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import init_logging, move_cursor_up
 from lerobot.utils.visualization_utils import init_rerun, log_rerun_data
 
@@ -173,7 +173,7 @@ def teleop_loop(
             move_cursor_up(len(robot_action_to_send) + 5)
 
         dt_s = time.perf_counter() - loop_start
-        busy_wait(1 / fps - dt_s)
+        precise_sleep(1 / fps - dt_s)
         loop_s = time.perf_counter() - loop_start
         print(f"\ntime: {loop_s * 1e3:.2f}ms ({1 / loop_s:.0f} Hz)")
 

src/lerobot/utils/robot_utils.py

@@ -16,14 +16,40 @@ import platform
 import time
 
 
-def busy_wait(seconds):
-    if platform.system() == "Darwin" or platform.system() == "Windows":
-        # On Mac and Windows, `time.sleep` is not accurate and we need to use this while loop trick,
-        # but it consumes CPU cycles.
+def precise_sleep(seconds: float, spin_threshold: float = 0.010, sleep_margin: float = 0.003):
+    """
+    Wait for `seconds` with better precision than time.sleep alone at the expense of more CPU usage.
+
+    Parameters:
+      - seconds: duration to wait
+      - spin_threshold: if remaining <= spin_threshold -> spin; otherwise sleep (seconds). Default 10ms
+      - sleep_margin: when sleeping leave this much time before deadline to avoid oversleep. Default 3ms
+
+    Note:
+        The default parameters are chosen to prioritize timing accuracy over CPU usage for the common 30 FPS use case.
+    """
+    if seconds <= 0:
+        return
+
+    system = platform.system()
+    # On macOS and Windows the scheduler / sleep granularity can make
+    # short sleeps inaccurate. Instead of burning CPU for the whole
+    # duration, sleep for most of the time and spin for the final few
+    # milliseconds to achieve good accuracy with much lower CPU usage.
+    if system in ("Darwin", "Windows"):
         end_time = time.perf_counter() + seconds
-        while time.perf_counter() < end_time:
-            pass
+        while True:
+            remaining = end_time - time.perf_counter()
+            if remaining <= 0:
+                break
+            # If there's more than a couple milliseconds left, sleep most
+            # of the remaining time and leave a small margin for the final spin.
+            if remaining > spin_threshold:
+                # Sleep but avoid sleeping past the end by leaving a small margin.
+                time.sleep(max(remaining - sleep_margin, 0))
+            else:
+                # Final short spin to hit precise timing without long sleeps.
+                pass
     else:
-        # On Linux time.sleep is accurate
-        if seconds > 0:
-            time.sleep(seconds)
+        # On Linux time.sleep is accurate enough for most uses
+        time.sleep(seconds)