Update 2_project.md

_projects/2_project.md

@@ -7,7 +7,7 @@ importance: 2
 category: LUT
 ---
 
-Since 2020, I have been pursuing a doctoral degree at the Intelligent Machinery Laborato-ry within the Department of Mechanical Engineering at LUT University in Finland. My su-pervisors are Prof. Heikki Handroos and Prof. Huapeng Wu.We are involved in the overarching research initiative known as ITER (International Ther-monuclear Experimental Reactor). Our primary responsibility lies in the remote handling maintenance operations within the reactor's vacuum chamber.
+Since 2020, I have been pursuing a doctoral degree at the Intelligent Machinery Laborato-ry within the Department of Mechanical Engineering at LUT University in Finland. My su-pervisors are Prof. Heikki Handroos and Prof. Huapeng Wu. We are involved in the overarching research initiative known as ITER (International Ther-monuclear Experimental Reactor). Our primary responsibility lies in the remote handling maintenance operations within the reactor's vacuum chamber.
 
 The environment within the reactor's vacuum chamber is considerably more challenging than ordinary settings, primarily due to its construction from smooth metallic materials. Depth cameras or LiDAR struggle to provide accurate observations in such conditions. Consequently, we need to implement various automated maintenance operations in the absence of reliable three-dimensional data.
 
@@ -34,7 +34,7 @@ My specific research focus comprises two components: robotic grasping and high-p
     The strucure of muti-sensor based peg-in-hole assembly method.
 </div>
 
-Following the robotic grasping task, we delved into high-precision rigid peg-in-hole assembly. In this study, the gap between the axis and hole was less than 0.1 millimeters (even lower than the repeatability accuracy of the UR10 robot we utilized), with a certain degree of randomness in the relative position between the end-effector and the grasped axis. To address these challenges, we fused data from a monocular camera and force sensors, employing deep reinforcement learning to train the robotic arm to execute assembly operations in a manner akin to human hand-eye coordination. This approach ultimately yielded excellent results. Some of the experiments is shown in the video.
+Following the robotic grasping task, we delved into high-precision rigid peg-in-hole assembly. In this study, the gap between the axis and hole was less than 0.08 millimeters (**lower than the repeatability accuracy of the UR10 robot we utilized**), with a certain degree of randomness in the relative position between the end-effector and the grasped axis. To address these challenges, we fused data from a monocular camera and force sensors, employing deep reinforcement learning to train the robotic arm to execute assembly operations in a manner akin to human hand-eye coordination. This approach ultimately yielded excellent results. Some of the experiments is shown in the video.
 
 <div class="row justify-content-sm-center">
     <div class="col-10">