diff --git a/nodes/simulation.py b/bt_sim_node.py
similarity index 72%
rename from nodes/simulation.py
rename to bt_sim_node.py
index 3cae5f1..e56a5b6 100755
--- a/nodes/simulation.py
+++ b/bt_sim_node.py
@@ -9,11 +9,13 @@ import rospy
 import franka_gripper.msg
 from geometry_msgs.msg import Pose
 from sensor_msgs.msg import JointState
+import std_srvs.srv
 
-from robot_tools.btsim import BtPandaEnv
 from robot_tools.controllers import CartesianPoseController
 from robot_tools.ros import *
 
+from simulation import BtPandaEnv
+
 CONTROLLER_UPDATE_RATE = 60
 JOINT_STATE_PUBLISHER_RATE = 60
 
@@ -22,26 +24,37 @@ class BtSimNode:
     def __init__(self, gui):
         self.sim = BtPandaEnv(gui=gui, sleep=False)
         self.controller = CartesianPoseController(self.sim.arm)
-
-        self.joint_state_pub = rospy.Publisher(
-            "/joint_states", JointState, queue_size=10
-        )
+        self.reset_server = rospy.Service("reset", std_srvs.srv.Trigger, self.reset)
         self.move_server = actionlib.SimpleActionServer(
             "move",
             franka_gripper.msg.MoveAction,
             execute_cb=self.move,
             auto_start=False,
         )
+        self.joint_state_pub = rospy.Publisher(
+            "joint_states", JointState, queue_size=10
+        )
+        rospy.Subscriber("target", Pose, self.target_pose_cb)
+        self.step_cnt = 0
+        self.reset_requested = False
         self.move_server.start()
-        rospy.Subscriber("/target", Pose, self.target_pose_cb)
+
+    def reset(self, req):
+        self.reset_requested = True
+        rospy.sleep(1.0)  # wait for the latest sim step to finish
+        self.sim.reset()
+        self.controller.set_target(self.sim.arm.pose())
+        self.step_cnt = 0
+        self.reset_requested = False
+        return std_srvs.srv.TriggerResponse(success=True)
 
     def run(self):
         rate = rospy.Rate(self.sim.rate)
-        self.step_cnt = 0
         while not rospy.is_shutdown():
-            self.handle_updates()
-            self.sim.step()
-            self.step_cnt = (self.step_cnt + 1) % self.sim.rate
+            if not self.reset_requested:
+                self.handle_updates()
+                self.sim.step()
+                self.step_cnt = (self.step_cnt + 1) % self.sim.rate
             rate.sleep()
 
     def handle_updates(self):
diff --git a/launch/simulation.launch b/launch/simulation.launch
index 1e1c92f..8acd595 100644
--- a/launch/simulation.launch
+++ b/launch/simulation.launch
@@ -7,7 +7,7 @@
   <param name="robot_description" command="$(find xacro)/xacro $(find franka_description)/robots/panda_arm_hand.urdf.xacro" />
 
   <!-- Bringup the robot -->
-  <node pkg="active_grasp" type="simulation.py" name="simulation" args="--gui" output="screen" />
+  <node pkg="active_grasp" type="bt_sim_node.py" name="simulation" args="--gui" output="screen" />
   <node pkg="robot_state_publisher" type="robot_state_publisher" name="robot_state_publisher" />
 
   <!-- Visualize the robot. -->
diff --git a/run.py b/run.py
index e7a12c6..af1fa46 100644
--- a/run.py
+++ b/run.py
@@ -1,12 +1,12 @@
 import argparse
 
-from geometry_msgs.msg import Pose
 import numpy as np
 import rospy
-from std_srvs.srv import Trigger
+
+from geometry_msgs.msg import Pose
+import std_srvs.srv
 
 from policies import get_policy
-
 from robot_tools.ros import *
 
 
@@ -14,16 +14,25 @@ class GraspController:
     def __init__(self, policy, rate):
         self.policy = policy
         self.rate = rate
-
-        self.target_pose_pub = rospy.Publisher("/target", Pose, queue_size=10)
+        self.reset_client = rospy.ServiceProxy("reset", std_srvs.srv.Trigger)
+        self.target_pose_pub = rospy.Publisher("target", Pose, queue_size=10)
         self.gripper = PandaGripperRosInterface()
+        rospy.sleep(1.0)
+
+    def run(self):
+        self.reset()
+        self.explore()
+        self.execute_grasp()
+
+    def reset(self):
+        req = std_srvs.srv.TriggerRequest()
+        self.reset_client(req)
 
     def explore(self):
         r = rospy.Rate(self.rate)
-        done = False
         self.policy.start()
-        while not done:
-            done = self.policy.update()
+        while not self.policy.done:
+            self.policy.update()
             r.sleep()
 
     def execute_grasp(self):
@@ -41,11 +50,9 @@ class GraspController:
 
 def main(args):
     rospy.init_node("panda_grasp")
-
     policy = get_policy(args.policy)
     gc = GraspController(policy, args.rate)
-    gc.explore()
-    gc.execute_grasp()
+    gc.run()
 
 
 if __name__ == "__main__":
diff --git a/simulation.py b/simulation.py
new file mode 100644
index 0000000..dea6ba8
--- /dev/null
+++ b/simulation.py
@@ -0,0 +1,35 @@
+import pybullet as p
+
+from robot_tools.btsim import *
+from robot_tools.spatial import Rotation, Transform
+
+
+class BtPandaEnv(BtBaseEnv):
+    def __init__(self, gui=True, sleep=True):
+        super().__init__(gui, sleep)
+        self.arm = BtPandaArm()
+        self.gripper = BtPandaGripper()
+        self.T_W_B = Transform(Rotation.identity(), np.r_[-0.6, 0.0, 0.4])
+        self.load_table()
+        self.load_robot()
+        self.load_objects()
+
+    def reset(self):
+        q = self.arm.configurations["ready"]
+        for i, q_i in enumerate(q):
+            p.resetJointState(self.arm.uid, i, q_i, 0)
+
+    def load_table(self):
+        p.loadURDF("plane.urdf")
+        p.loadURDF(
+            "table/table.urdf",
+            baseOrientation=Rotation.from_rotvec(np.array([0, 0, np.pi / 2])).as_quat(),
+            useFixedBase=True,
+        )
+
+    def load_robot(self):
+        self.arm.load(self.T_W_B)
+        self.gripper.uid = self.arm.uid
+
+    def load_objects(self):
+        p.loadURDF("cube_small.urdf", [-0.2, 0.0, 0.8])