1. Introduction:

This text primarily introduces the sensible utility of LIMO Cobot by Elephant Robotics in a simulated situation. You might have seen earlier posts about LIMO Cobot’s technical instances, A[LINK], B[LINK]. The explanation for writing one other associated article is that the unique testing setting, whereas demonstrating fundamental performance, typically seems overly idealized and simplified when simulating real-world functions. Subsequently, we intention to make use of it in a extra operationally constant setting and share a few of the points that arose at the moment.

2. Evaluating the Previous and New Eventualities:

First, let’s take a look at what the previous and new situations are like.

Previous State of affairs: A easy setup with just a few obstacles, comparatively common objects, and a subject enclosed by obstacles, roughly 1.5m*2m in dimension.

New State of affairs: The brand new situation comprises a greater diversity of obstacles of various shapes, together with a hollowed-out object within the center, simulating an actual setting with street steering markers, parking areas, and extra. The scale of the sphere is 3m*3m.

The change in setting is important for testing and demonstrating the comprehensiveness and applicability of our product.

3. Evaluation of Sensible Circumstances:

Subsequent, let’s briefly introduce the general course of.

The method is principally divided into three modules: one is the performance of LIMO PRO, the second is machine imaginative and prescient processing, and the third is the performance of the robotic arm. (For a extra detailed introduction, please see the earlier article https://robots-blog.com/2024/05/16/exploring-elephant-robotics-limo-cobot/.)

LIMO PRO is principally chargeable for SLAM mapping, utilizing the gmapping algorithm to map the terrain, navigate, and in the end obtain the operate of fixed-point patrol.

myCobot 280 M5 is primarily chargeable for the duty of greedy objects. A digicam and a suction pump actuator are put in on the finish of the robotic arm. The digicam captures the true scene, and the picture is processed by the OpenCV algorithm to search out the coordinates of the goal object and carry out the greedy operation.

Total course of:

1. LIMO performs mapping.⇛

2. Run the fixed-point cruising program.⇛

3. LIMO goes to level A ⇛ myCobot 280 performs the greedy operation ⇒ goes to level B ⇛ myCobot 280 performs the inserting operation.

4. ↺ Repeat step 3 till there are not any goal objects, then terminate this system.

Subsequent, let’s observe the sensible execution course of.

Mapping:

First, you must begin the radar by opening a brand new terminal and getting into the next command:

roslaunch limo_bringup limo_start.launch pub_odom_tf:=false

Then, begin the gmapping mapping algorithm by opening one other new terminal and getting into the command:

roslaunch limo_bringup limo_gmapping.launch

After profitable startup, the rviz visualization software will open, and you will notice the interface as proven within the determine.

At this level, you possibly can change the controller to distant management mode to regulate the LIMO for mapping.

After setting up the map, you must run the next instructions to save lots of the map to a specified listing:

1. Swap to the listing the place you need to save the map. Right here, save the map to `~/agilex_ws/src/limo_ros/limo_bringup/maps/`. Enter the command within the terminal:

cd ~/agilex_ws/src/limo_ros/limo_bringup/maps/

2. After switching to `/agilex_ws/limo_bringup/maps`, proceed to enter the command within the terminal:

rosrun map_server map_saver -f map1

This course of went very easily. Let’s proceed by testing the navigation operate from level A to level B.

Navigation:

1. First, begin the radar by getting into the next command within the terminal:

roslaunch limo_bringup limo_start.launch pub_odom_tf:=false

2. Begin the navigation operate by getting into the next command within the terminal:

roslaunch limo_bringup limo_navigation_diff.launch

Upon success, this interface will open, displaying the map we simply created.

Click on on „2D Pose Estimate, “ then click on on the placement the place LIMO is on the map. After beginning navigation, you will discover that the form scanned by the laser doesn’t overlap with the map. You must manually right this by adjusting the precise place of the chassis within the scene on the map displayed in rviz. Use the instruments in rviz to publish an approximate place for LIMO. Then, use the controller to rotate LIMO, permitting it to auto-correct. When the form of the laser scan overlaps with the shapes within the map’s scene, the correction is full, as proven within the determine the place the scanned form and the map overlap.

Click on on „2D Nav Aim“ and choose the vacation spot on the map for navigation.

The navigation check additionally proceeds easily.

Subsequent, we’ll transfer on to the half in regards to the static robotic arm’s greedy operate.

Figuring out and Buying the Pose of Aruco Codes

To exactly determine objects and procure the place of the goal object, we processed Aruco codes. Earlier than beginning, guarantee the particular parameters of the digicam are set.

Initialize the digicam parameters primarily based on the digicam getting used.

def __init__(self, mtx: np.ndarray, dist: np.ndarray, marker_size: int):
self.mtx = mtx
self.dist = dist
self.marker_size = marker_size
self.aruco_dict = cv2.aruco.Dictionary_get(cv2.aruco.DICT_6X6_250)
self.parameters = cv2.aruco.DetectorParameters_create()

Then, determine the thing and estimate its pose to acquire the 3D place of the thing and output the place info.

def estimatePoseSingleMarkers(self, corners):
"""
This may estimate the rvec and tvec for every of the marker corners detected by:
corners, ids, rejectedImgPoints = detector.detectMarkers(picture)
corners - is an array of detected corners for every detected marker within the picture
marker_size - is the dimensions of the detected markers
mtx - is the digicam matrix
distortion - is the digicam distortion matrix
RETURN record of rvecs, tvecs, and trash (in order that it corresponds to the previous estimatePoseSingleMarkers())
"""
marker_points = np.array([[-self.marker_size / 2, self.marker_size / 2, 0],
[self.marker_size / 2, self.marker_size / 2, 0],
[self.marker_size / 2, -self.marker_size / 2, 0],
[-self.marker_size / 2, -self.marker_size / 2, 0]], dtype=np.float32)
rvecs = []
tvecs = []
for nook in corners:
retval, rvec, tvec = cv2.solvePnP(marker_points, nook, self.mtx, self.dist, False,
cv2.SOLVEPNP_IPPE_SQUARE)
if retval:
rvecs.append(rvec)
tvecs.append(tvec)
rvecs = np.array(rvecs)
tvecs = np.array(tvecs)
(rvecs - tvecs).any()
return rvecs, tvecs

The steps above full the identification and acquisition of the thing’s info, and eventually, the thing’s coordinates are returned to the robotic arm to execute the greedy.

Robotic Arm Motion and Greedy Operation

Primarily based on the place of the Aruco marker, calculate the goal coordinates the robotic arm wants to maneuver to and convert the place right into a coordinate system appropriate for the robotic arm.

def homo_transform_matrix(x, y, z, rx, ry, rz, order="ZYX"):
rot_mat = rotation_matrix(rx, ry, rz, order=order)
trans_vec = np.array([[x, y, z, 1]]).T
mat = np.vstack([rot_mat, np.zeros((1, 3))])
mat = np.hstack([mat, trans_vec])
return mat

If the Z-axis place is detected as too excessive, will probably be corrected:

if end_effector_z_height shouldn't be None:  
p_base[2] = end_effector_z_height

After the coordinate correction is accomplished, the robotic arm will transfer to the goal place.

# Concatenate x, y, z, and the present posture into a brand new array
new_coords = np.concatenate([p_base, curr_rotation[3:]]) 
xy_coords = new_coords.copy()

Then, management the top effector’s API to suction the thing.

The above completes the respective features of the 2 robots. Subsequent, they are going to be built-in into the ROS setting.

#Initialize the coordinates of level A and B
    goal_1 = [(2.060220241546631,-2.2297520637512207,0.009794792000444471,0.9999520298742676)] #B
    goal_2 = [(1.1215190887451172,-0.002757132053375244,-0.7129997613218174,0.7011642748707548)] #A
    #Begin navigation and hyperlink the robotic arm
    map_navigation = MapNavigation()
    arm = VisualGrasping("10.42.0.203",9000)
    print("join profitable")

    arm.perform_visual_grasp(1,-89)
    # Navigate to location A and carry out the duty
        for purpose in goal_1:
        x_goal, y_goal, orientation_z, orientation_w = purpose
        flag_feed_goalReached = map_navigation.moveToGoal(x_goal, y_goal, orientation_z, orientation_w)
        if flag_feed_goalReached:
            time.sleep(1)
            # executing 1 seize and setting the top effector's Z-axis peak to -93.
            arm.unload()
            print("command accomplished")
        else:
            print("failed")

4. Issues Encountered

Mapping Scenario:

After we initially tried mapping with out enclosing the sphere, frequent errors occurred throughout navigation and localization, and it failed to satisfy our necessities for a simulated situation.

Navigation Scenario:

Within the new situation, one of many obstacles has a hole construction.

Throughout navigation from level A to level B, LIMO could fail to detect this impediment and assume it may possibly go by, damaging the unique impediment. This concern arises as a result of LIMO’s radar is positioned low, scanning solely the empty house. Potential options embody adjusting the radar’s scanning vary, which requires in depth testing for fine-tuning, or adjusting the radar’s peak to make sure the impediment is acknowledged as impassable.

Robotic Arm Greedy Scenario:

Within the video, it’s evident that our goal object is positioned on a flat floor. The greedy didn’t think about impediment avoidance for the thing. Sooner or later, when setting particular positions for greedy, this case must be thought-about.

5. Conclusion

Total, LIMO Cobot carried out excellently on this situation, efficiently assembly the necessities. Your entire simulated situation coated a number of core areas of robotics, together with movement management of the robotic arm, path planning, machine imaginative and prescient recognition and greedy, and radar mapping navigation and fixed-point cruising features of the cellular chassis. By integrating these practical modules in ROS, we constructed an environment friendly automated course of, showcasing LIMO Cobot’s broad adaptability and superior capabilities in complicated environments.

Credit

Elephant Robotics