diff --git a/datasets/intel_lasers.npy b/datasets/intel_lasers.npy
index bca7ef5..3af808b 100644
Binary files a/datasets/intel_lasers.npy and b/datasets/intel_lasers.npy differ
diff --git a/icp.py b/icp.py
index 7b1f73d..2920e82 100644
--- a/icp.py
+++ b/icp.py
@@ -158,13 +158,6 @@ def icp(A, B, init_pose=None, max_iterations=20, tolerance=0.001):
         i: number of iterations to converge
     '''
 
-    size = np.min([A.shape[0], B.shape[0]])
-    prev_random_idx = np.sort(np.random.choice(
-        np.arange(A.shape[0]), size, replace=True))
-    cur_random_idx = np.sort(np.random.choice(
-        np.arange(B.shape[0]), size, replace=True))
-    A = A[prev_random_idx]
-    B = B[cur_random_idx]
     assert A.shape == B.shape
 
     # get number of dimensions
@@ -187,13 +180,16 @@ def icp(A, B, init_pose=None, max_iterations=20, tolerance=0.001):
         distances, indices = nearest_neighbor(src[:m,:].T, dst[:m,:].T)
 
         # Reject pairs that have 1 meter distance between them
+        indices = indices[np.linalg.norm(src[:m, :], axis=0) < 80]
+        distances = distances[np.linalg.norm(src[:m, :], axis=0) < 80]
+        filtered_src = src[:, np.linalg.norm(src[:m, :], axis=0) < 80]
         indices = indices[distances < 1.0]
 
         # compute the transformation between
         # the current source and nearest destination points
-        T,_,_ = best_fit_transform(src[:m, distances < 1.0].T, dst[:m,indices].T)
+        T,_,_ = best_fit_transform(filtered_src[:m, distances < 1.0].T, dst[:m,indices].T)
 
-        distances = distances[distances <= 1.0]
+        # distances = distances[distances <= 1.0]
 
         # update the current source
         src = np.dot(T, src)
@@ -207,11 +203,11 @@ def icp(A, B, init_pose=None, max_iterations=20, tolerance=0.001):
     # calculate final transformation
     T,_,_ = best_fit_transform(A, src[:m,:].T)
 
-    theta = np.arctan2(T[1,0], T[0,0])
-    t = T[0:2, 2]
-    angle_res = 1.0
 
     # FIXME:
+    # theta = np.arctan2(T[1,0], T[0,0])
+    # t = T[0:2, 2]
+    # angle_res = 1.0
     # angles = cur_random_idx * angle_res - 90 
     # cov, _, _ = compute_covariance(src[:m, :].T, A, t, theta, np.radians(angles))
     cov = np.eye(3)
diff --git a/slam.py b/slam.py
index 82b27b1..8056631 100644
--- a/slam.py
+++ b/slam.py
@@ -33,9 +33,6 @@ def eigsorted(cov):
 
 parser = argparse.ArgumentParser(description='Python Graph Slam')
 
-parser.add_argument('--seed', default=None, type=int,
-                    help='Random number generator seed')
-
 parser.add_argument('--draw_last', default=float('inf'), type=int,
                     help='Number of point clouds to draw.')
 
@@ -46,16 +43,13 @@ def eigsorted(cov):
 parser.set_defaults(save_gif=False)
 
 args = parser.parse_args()
-    
+
 if args.save_gif:
     import atexit
     images = []
     atexit.register(lambda: imageio.mimsave(f'./slam_{int(time.time())}.gif',
                                             images, fps=10))
 
-if args.seed is not None:
-    np.random.seed(args.seed) # For testing
-
 # Starting point 
 optimizer = pose_graph.PoseGraphOptimization()
 pose = np.eye(3)
@@ -78,7 +72,7 @@ def eigsorted(cov):
         continue
 
     dx = odom - prev_odom
-    if np.linalg.norm(dx[0:2]) > 0.3 or abs(dx[2]) > 0.2:
+    if np.linalg.norm(dx[0:2]) > 0.5 or abs(dx[2]) > 0.2:
         # Scan Matching
         A = lasers[prev_idx]
         B = lasers[odom_idx]
@@ -133,8 +127,8 @@ def eigsorted(cov):
                                 max_iterations=80, tolerance=0.0001)
                         except Exception as e:
                             continue
-                    information = np.linalg.inv(cov) 
-                    if np.mean(distances) < 0.05:
+                    information = np.linalg.inv(cov)
+                    if np.mean(distances) < 0.2:
                         rk = g2o.RobustKernelDCS()
                         optimizer.add_edge([vertex_idx, idx],
                                            g2o.SE2(g2o.Isometry2d(tran)),
@@ -153,7 +147,9 @@ def eigsorted(cov):
             x = optimizer.get_pose(idx)
             r = x.to_isometry().R
             t = x.to_isometry().t
-            point_cloud.append((r @ registered_lasers[idx].T + t[:, np.newaxis]).T)
+            filtered = registered_lasers[idx]
+            filtered = filtered[np.linalg.norm(filtered, axis=1) < 80]
+            point_cloud.append((r @ filtered.T + t[:, np.newaxis]).T)
             traj.append(x.to_vector()[0:2])
         point_cloud = np.vstack(point_cloud)