Patent Application
20190155295
The invention described herein relates to an electrical charging system and more particularly to the navigation of a robot to an electrical charging system and the docking of the robot to the electrical charging system.
In many applications, robots are used to perform functions in place of humans or to assist humans in order to increase productivity and efficiency. One such application is order fulfillment, which is typically performed in a large warehouse filled with products to be shipped to customers who have placed their orders over the internet for home delivery. Fulfilling such orders in a timely, accurate and efficient manner is logistically challenging to say the least.
In an online Internet shopping application, for example, clicking the “check out” button in a virtual shopping cart creates an “order.” The order includes a listing of items that are to be shipped to a particular address. The process of “fulfillment” involves physically taking or “picking” these items from a large warehouse, packing them, and shipping them to the designated address.
An important goal of the order fulfillment process is thus to ship as many items in as short a time as possible. The process of receiving an order, planning its fulfillment, finding the storage shelf or bin, picking the product, and repeating the process for each item on the order, then delivering the order to a shipping station is repetitive and labor intensive. In a warehouse stocked with thousands or tens of thousands of items of rapidly turning inventory, robots play a critical role in ensuring timely and efficient order fulfillment. In addition, the products that will ultimately be shipped first need to be received in the warehouse and stored or “placed” in storage bins in an orderly fashion throughout the warehouse so they can be readily retrieved for shipping.
Using robots to perform picking and placing functions may be done by the robot alone or with the assistance of human operators. Picking and placing or stocking functions, whether or not performed with human interaction, requires that the robot navigate from its present location to a target product storage or “bin” location. One method of navigation by a robot in an order fulfilment warehouse employs a spatial model or “map” of the warehouse, locally stored and updated by the robot, to allow the robot to operate autonomously or semi-autonomously as it performs its assigned order fulfillment tasks. The map is a digital representation of the warehouse, its storage locations, obstacles and other features. To arrive at a product bin in the presence of stationary and dynamic obstacles, the robot performs processing operations on the map to determine its present location and for continually recalibrating its movement along the goal path.
The robots are powered by electricity, which is stored in batteries onboard the robot. With all of the travelling that the robots do around the warehouse they must be regularly recharged. Therefore, for the operation to run smoothly, an efficient and effective way to charge the robots is a requirement. For general navigation within a warehouse, the size and resolution of the map may be such that a robot can successfully navigate to its target location, while avoiding obstacles along its goal path. Processing on the warehouse map, however, may require too much processing and result in too coarse of a localization and control where more precise localization and control is needed, such as when docking the robot to a robot charging station.
What is needed is a computationally efficient approach to localizing and controlling the robot during the docking of a robot to a robot charging station.
The benefits and advantages of the present invention over existing systems will be readily apparent from the Brief Summary of the Invention and Detailed Description to follow. One skilled in the art will appreciate that the present teachings can be practiced with embodiments other than those summarized or disclosed below.
In one aspect of the invention, there is a method for navigating a robot for docking with a charger docking station. The method may be include receiving an initial pose associated with a robot charger docking station, receiving a mating pose associated with the robot charger docking station, performing a first navigation of a robot from a current pose to the initial pose, and performing a second navigation of the robot from the initial pose to the mating pose. The second navigation of the robot may proceed substantially along an arc path from the initial pose to the mating pose, thereby, upon arriving at the mating pose, an electrical charging port of the robot mates with an electrical charging assembly of the charger docking station.
In one embodiment of the first aspect, the arc path from the initial pose to the mating pose may be associated with a section of a unique circle having a radius and a center equidistant from a first location associated with the initial pose and a second location associated with the mating pose. In some embodiments, an instantaneous linear velocity and an instantaneous angular velocity of the robot at a pose along the arc path are maintained in substantially constant relation with the radius. Controlling a rotational error of the robot using such relation may include issuing a proportional control and/or a weighted control. In further embodiments, weighting parameters of the weighted control may be adjusted in nonlinear relation as a function of the distance to the charging station.
In the above embodiments, a method of controlling the error of the robot along the arc path may include closing the rotational error with a proportional control until the error gets below a threshold, setting a linear velocity to a fixed value, controlling the robot according to a weighted control, constantly updating a radius and rotational error, and when a threshold is exceeded, switching the control from the weighted control to the proportional control until the error returns to below the threshold.
In a second aspect of the invention, there is a mobile robot configured to navigate from a current location to and dock with a charger docking station for re-charging. The mobile robot may include a wheeled mobile base having an electrical charging port and a processor. The processor of the wheeled mobile base may be configured to obtain an initial pose associated with the charger docking station, obtain a mating pose associated with the charger docking station, navigate the wheeled mobile base from the current location to the initial pose, and navigate the wheeled mobile base from the initial pose to the mating pose. In some embodiments, the wheeled mobile base may proceed substantially along an arc path from the initial pose to the mating pose, thereby causing the electrical charging port of the wheeled base to mate with an electrical charging assembly of the robot charger station.
In an embodiment of the second aspect, the arc path from the initial pose to the mating pose may be associated with a section of a unique circle having a radius and a center equidistant from a first location associated with the initial pose and a second location associated with the mating pose. In some embodiments, an instantaneous linear velocity and an instantaneous angular velocity of the robot at a pose along the arc path are maintained in substantially constant relation with radius. Controlling a rotational error of the robot using such relation may include issuing a proportional control and/or a weighted control. In further embodiments, weighting parameters of the weighted control may be adjusted in nonlinear relation as a function of the distance to the charging station.
In the above embodiments, controlling the error of the robot along the arc path may include closing the rotational error with a proportional control until the error gets below a threshold, setting a linear velocity to a fixed value, controlling the robot according to a weighted control, constantly updating a radius and a rotational error, and when a threshold is exceeded, switching the control from the weighted control to the proportional control until the error returns to below the threshold.
In a third aspect, there is a robot system including a transceiver, a data processor and a data storage device having instructions stored thereon for execution by the data processor. The instructions for execution by the processor may be configured to receive an initial pose associated with a robot charger docking station, receive a mating pose associated with the robot charger docking station, perform a first navigation of a robot from a current pose to the initial pose, and perform a second navigation of the robot from the initial pose to the mating pose. In some embodiments, the second navigation may proceed substantially along an arc path from the initial pose to the mating pose, thereby causing the electrical charging port of the wheeled base to mate with an electrical charging assembly of the robot charger station.
In an embodiment of the third aspect, the arc path from the initial pose to the mating pose may be associated with a section of a unique circle having a radius and a center equidistant from a first location associated with the initial pose and a second location associated with the mating pose. In some embodiments, an instantaneous linear velocity and an instantaneous angular velocity of the robot, at a pose along the arc path, are maintained in substantially constant relation with the radius. Controlling a rotational error of the robot using such relation may include issuing a proportional control and/or a weighted control.
In further embodiments, the weighting parameters of the weighted control may be adjusted in nonlinear relation as a function of the distance to the charging station. In the above embodiments, controlling the error of the robot along the arc path may include closing the rotational error with a proportional control until the error gets below a threshold, setting a linear velocity to a fixed value, controlling the robot according to a weighted control, constantly updating a radius and a rotational error, and when a threshold is exceeded, switching the control from the weighted control to the proportional control until the error returns to below the threshold.
These and other features of the invention will be apparent from the following detailed description and the accompanying figures.
Embodiments of the present invention will now be described, by way of example only, with reference to the attached Figures, wherein:
The disclosure and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments and examples that are described and/or illustrated in the accompanying drawings and detailed in the following description. It should be noted that the features illustrated in the drawings are not necessarily drawn to scale, and features of one embodiment may be employed with other embodiments as the skilled artisan would recognize, even if not explicitly stated herein. Descriptions of well-known components and processing techniques may be omitted so as to not unnecessarily obscure the embodiments of the disclosure. The examples used herein are intended merely to facilitate an understanding of ways in which the disclosure may be practiced and to further enable those of skill in the art to practice the embodiments of the disclosure. Accordingly, the examples and embodiments herein should not be construed as limiting the scope of the disclosure. Moreover, it is noted that like reference numerals represent similar parts throughout the several views of the drawings.
The invention is directed to the docking of robots to an electrical charging system. Although not restricted to any particular robot application, one suitable application that the invention may be used in is order fulfillment. The use of robots in this application will be described to provide context for docking the robot to an electrical charging system.
While the description provided herein is focused on picking items from bin locations in the warehouse to fulfill an order for shipment to a customer, the system is equally applicable to the storage or placing of items received into the warehouse in bin locations throughout the warehouse for later retrieval and shipment to a customer. The invention is also applicable to inventory control tasks associated with such a warehouse system, such as, consolidation, counting, verification, inspection and clean-up of products. These and other benefits and advantages will become readily apparent from the examples and illustrations described below.
Referring to
In a preferred embodiment, a robot 18, shown in
Referring again to
Although a robot 18 excels at moving around the warehouse 10, with current robot technology, it is not very good at quickly and efficiently picking items from a shelf and placing them in the tote 44 due to the technical difficulties associated with robotic manipulation of objects. A more efficient way of picking items is to use a local operator 50, which is typically human, to carry out the task of physically removing an ordered item from a shelf 12 and placing it on robot 18, for example, in tote 44. The robot 18 communicates the order to the local operator 50 via the tablet 48 (or laptop/other user input device), which the local operator 50 can read, or by transmitting the order to a handheld device used by the local operator 50.
Upon receiving an order 16 from the order server 14, the robot 18 proceeds to a first warehouse location, e.g. as shown in
Upon reaching the correct location, the robot 18 parks itself in front of a shelf 12 on which the item is stored and waits for a local operator 50 to retrieve the item from the shelf 12 and place it in tote 44. If robot 18 has other items to retrieve it proceeds to those locations. The item(s) retrieved by robot 18 are then delivered to a packing station 100,
It will be understood by those skilled in the art that each robot may be fulfilling one or more orders and each order may consist of one or more items. Typically, some form of route optimization software would be included to increase efficiency, but this is beyond the scope of this invention and is therefore not described herein.
In order to simplify the description of the invention, a single robot 18 and operator 50 are described. However, as is evident from
The navigation approach of this invention, as well as the semantic mapping of a SKU of an item to be retrieved to a fiducial ID/pose associated with a fiducial marker in the warehouse where the item is located, is described in detail below with respect to
Using one or more robots 18, a map of the warehouse 10 must be created and dynamically updated to determine the location of objects, both static and dynamic, as well as the locations of various fiducial markers dispersed throughout the warehouse. To do this, one of the robots 18 navigate the warehouse and build/update a map 10a,
Robot 18 utilizes its laser-radar 22 to create/update map 10a of warehouse 10 as robot 18 travels throughout the space identifying open space 112, walls 114, objects 116, and other static obstacles such as shelves 12a in the space, based on the reflections it receives as the laser-radar scans the environment.
While constructing the map 10a or thereafter, one or more robots 18 navigates through warehouse 10 using cameras 24a and 24b to scan the environment to locate fiducial markers (two-dimensional bar codes) dispersed throughout the warehouse on shelves proximate bins, such as 32 and 34,
By the use of wheel encoders and heading sensors, vector 120, and the robot's position in the warehouse 10 can be determined. Using the captured image of a fiducial marker/two-dimensional barcode and its known size, robot 18 can determine the orientation with respect to and distance from the robot of the fiducial marker/two-dimensional barcode, vector 130. With vectors 120 and 130 known, vector 140, between origin 110 and fiducial marker 30, can be determined. From vector 140 and the determined orientation of the fiducial marker/two-dimensional barcode relative to robot 18, the pose (position and orientation) defined by a quaternion (x, y, z, ω) for fiducial marker 30 can be determined.
Flowchart 200,
In look-up table 300, which may be stored in the memory of each robot, there are included for each fiducial marker a fiducial identification, 1, 2, 3, etc., and a pose for the fiducial marker/bar code associated with each fiducial identification. The pose consists of the x,y,z coordinates in the warehouse along with the orientation or the quaternion (x,y,z, ω).
In another look-up Table 400,
The alpha-numeric bin locations are understandable to humans, e.g. operator 50,
The order fulfillment process according to this invention is depicted in flowchart 500,
Item specific information, such as SKU number and bin location, obtained by the warehouse management system 15, can be transmitted to tablet 48 on robot 18 so that the operator 50 can be informed of the particular items to be retrieved when the robot arrives at each fiducial marker location.
With the SLAM map and the pose of the fiducial ID's known, robot 18 can readily navigate to any one of the fiducial ID's using various robot navigation techniques. The preferred approach involves setting an initial route to the fiducial marker pose given the knowledge of the open space 112 in the warehouse 10 and the walls 114, shelves (such as shelf 12) and other obstacles 116. As the robot begins to traverse the warehouse using its laser radar 22, it determines if there are any obstacles in its path, either fixed or dynamic, such as other robots 18 and/or operators 50, and iteratively updates its path to the pose of the fiducial marker. The robot re-plans its route about once every 50 milliseconds, constantly searching for the most efficient and effective path while avoiding obstacles.
Generally, localization of the robot within warehouse 10a is achieved by many-to-many multiresolution scan matching (M3RSM) operating on the SLAM virtual map. Compared to brute force methods, M3RSM dramatically reduces the computational time for a robot to perform SLAM loop closure and scan matching, two critical steps in determining robot pose and position. Robot localization is further improved by minimizing the M3SRM search space according to methods disclosed in related U.S. application Ser. No. 15/712,222, entitled MULTI-RESOLUTION SCAN MATCHING WITH EXCLUSION ZONES, filed on Sep. 22, 2017, and incorporated by reference in its entirety herein.
With the product SKU/fiducial ID to fiducial pose mapping technique combined with the SLAM navigation technique both described herein, robots 18 are able to very efficiently and effectively navigate the warehouse space without having to use more complex navigation approaches typically used which involve grid lines and intermediate fiducial markers to determine location within the warehouse.
Generally, navigation in the presence of other robots and moving obstacles in the warehouse is achieved by collision avoidance methods including the dynamic window approach (DWA) and optimal reciprocal collision avoidance (ORCA). DWA computes among feasible robot motion trajectories an incremental movement that avoids collisions with obstacles and favors the desired path to the target fiducial marker. ORCA optimally avoids collisions with other moving robots without requiring communication with the other robot(s). Navigation proceeds as a series of incremental movements along trajectories computed at the approximately 50 ms update intervals. Collision avoidance may be further improved by techniques described in related U.S. application Ser. No. 15/712,256, entitled DYNAMIC WINDOW APPROACH USING OPTIMAL RECIPROCAL COLLISION AVOIDANCE COST-CRITIC, filed on Sep. 22, 2017, and incorporated by reference in its entirety herein.
As described above, robots 50 need to be periodically re-charged. In addition to marking locations in the warehouse where items are stored, a fiducial marker may be placed at one or more electrical charging station(s) within the warehouse. When robot 18 is low on power it can navigate to a fiducial marker located at an electrical charging station so it can be recharged. Once there it can be manually recharged by having an operator connect the robot to the electrical charging system or the robot can use its navigation to dock itself at the electrical charging station.
As shown in
First male terminal member 204 has first base 210 affixed to and extending orthogonally along a first axis 212 from surface 214 of the charger base 202 and terminates in a first electrical contact 216. First electrical contact 216 may be in the form of a copper bus bar which extends into charger base 202 to which would be affixed one of the positive or negative electrical connections. Second male terminal member 206 has second base 220 affixed to and extending orthogonally along a second axis 222 from surface 214 of the charger base 202 and terminates in a second electrical contact 226. Second electrical contact 226 may also be in the form of a copper bus bar which extends into charger base 202 to which would be affixed the other of the positive or negative electrical connections.
The first male terminal member 204 has a plurality of external surfaces at least two of which have a curved shape from the first base 210 to the first electrical contact 216 forming a concave surface. In the embodiment depicted in
In addition, first male terminal member 204 has a flat surface 236 which is substantially parallel to first axis 212 and orthogonal to surface 214 of charger base 202. Flat surface 236 includes a recessed surface portion 238 proximate first electrical contact 216.
The second male terminal member 206 has a plurality of external surfaces at least two of which have a curved shape from the second base 220 to the second electrical contact 226, forming a concave surface. In the embodiment depicted in
In addition, second male terminal member 206 has a flat surface 246, which is substantially parallel to second axis 222 and orthogonal to surface 214 of charger base 202. Flat surface 246 includes a flared surface portion 248 proximate second electrical contact 226.
There is a cavity 250 formed between the first male terminal member 204 and the second male terminal member 206 defined by the at least one flat surface 236 of the first male terminal member 204 and the at least one flat surface 246 of the second male terminal member 206. Cavity 250 has an opening 252 between the first electrical contact 216 and the second electrical contact 226. At opening 252, the recessed surface portion 238 of flat surface 236 and the flared surface portion 248 of flat surface 246, are present.
Referring again to
If the robot is docking to a fixed electrical charging station, it may use camera 24a and 24b to maneuver it into position so that electrical charging port 300 can mate with electrical charging assembly 200. The cameras may use the fiducial markers associated with the charging station as a reference point for fine localization, which will be described in more detail below. As the robot maneuvers into place, achieving perfect alignment for mating of the electrical contacts 216 and 226 of the electrical assembly 200 with electrical contacts 304 and 306, respectively, of electrical charging port 300 can be difficult. Therefore, electrical charging assembly 200 and electrical charging port 300 have been specifically designed in order to ensure easier, more efficient, and less problematic mating to allow the robots to electrically re-charge more quickly.
As can be seen in
The openings of cavities 308 and 310 are wider and longer than the width/length of the electrical contacts 216/226 of first male terminal member 204 second male terminal member 206. The extra width/length allows the first male terminal member 204 second male terminal member 206 to be more easily received within cavities 308 and 310 even if they are somewhat misaligned in the horizontal/vertical directions during the mating process. As the robot moves toward electrical charging assembly 200, the engagement of the complimentarily curved surfaces cause the first male terminal member 204 and the second male terminal member 206 to be guided into alignment so that engagement between electrical contacts 216/226 of electrical charging assembly and electrical contacts 304/306 of electrical charging port 300 will occur.
Thus, the radii of mating parts (male terminal members and cavities) are designed to provide coarse alignment when the male terminal members are first inserted into the cavities, and fine adjustment as full insertion is approached.
The electrical charging system provides an additional feature for easier vertical alignment. This is accomplished by the interaction of divider 320, which is between cavities 308 and 310, in combination with opening 352 of cavity 350 of electrical charging assembly 200. Flared surface portion 248 provides a wider opening so, if there is vertical misalignment, it causes the divider 320 to ride up vertically into place in cavity 350, as the docking process occurs.
When the first and second male terminals 204 and 206 are fully inserted into cavities 308 and 310, electrical charging assembly 200 is secured in place with electrical charging port 300 by means of magnets 360a-e, which engage with metal contacts 260a-e on electrical charging assembly 200. The magnets may be disposed beneath the external surface of electrical charging port 300 and, as such, they are shown in phantom.
There is an additional feature included in the electrical charging system, which is useful in the case of manual charging by an operator. If the electrical charging assembly 200 were inserted into the electrical charging port 300 improperly, i.e. upside down with electrical contact 216 of electrical charging assembly 200 connected to electrical contacts 306 of electrical charging port 300 and with electrical contact 226 of electrical charging assembly connected to electrical contacts 304 of electrical charging port 300, the polarities would be reversed and significant damage to robot base 20a would result.
To prevent this from happening, a stop 330 (see
As shown in
When electrical contacts 304 and 306 are in the compressed position, magnets 360a-e of electrical charging port 300 are in close proximity with metal contacts 260a-e of electrical charging assembly 200 and they magnetically engage to secure in place electrical charging assembly 200 and electrical charging port 300. In this position, it can be seen that upper and lower curved surfaces 230 and 240 of male terminal members 204 and 206, respectively, are complimentarily engaged with surfaces 312 and 314 of cavities 308 and 310, respectively.
Also depicted in
A charger docking station 500 according to an aspect of this invention is depicted in
Also shown is protective bumper 508, which may be made of metal, mounted horizontally across the bottom portion of front cover 502 to protect the charger docking station 500 from damage in the event that a robot does not smoothly dock. Charger docking station 500 further includes right side cover 510 and left side cover 512 (not visible in
A metal frame comprising front frame member 520a, right side frame member 520b, left side frame member 520c, and back side frame member 520d are interconnected to form the base structure for charger docking station 500. Referring to
Top cover 524, which is also made of a hard plastic material, includes a user interface panel 526 disposed in a cavity in the surface of top cover 524 which may include certain indicators and controls for a user to operate the charger docking station. For example, lighting signals to indicate various states such as “Ready”, “Charging”, “Power On”, “Recovery Mode”, and “Fault” or “E-Stop” may be included. Buttons such as “Power on/off”, “Start manual charge”, “Undock”, “Reset”, and “E-Stop” may be included.
Along the back edge of top cover 524 is a back panel 528, which comprises a center panel section 530 and side panel sections 532 and 534 on the right and left sides, respectively, of center panel 530. Center panel 530 has a rectangular front surface 536 which is substantially parallel to front cover 502. Right side panel 532 has a rectangular front surface 538 and left side panel 534 has a rectangular front surface 540.
Right and left side panels 532 and 534 have wide sidewalls 542 and 544, respectively, on one side and converge to narrower widths on the other sides which interconnect with center panel section 530. Thus, right and left side panels 532 and 534 are wedge-shaped. As a result, their front surfaces 538 and 540 are not parallel with front surface 536 of center panel 530 or front cover 502. They are each disposed at an angle, θ, with respect to surface 536. Fiducial markers 546 and 548 (e.g. a two-dimensional bar code) disposed on front surfaces 538 and 540, respectively, are also disposed at the angle, θ, relative to front surface 536 and the front cover 502.
As will be described in detail below, in one aspect the robots may use the angled fiducial markers for precision navigation during the process of docking with the charger docking station by viewing them with their onboard cameras. To generally navigate to the charger docking station when recharging is needed, the robots navigate in the same manner as they do when navigating to product bins as described above. Charging station 500 may be associated with a pose located in close proximity to the front cover 502 and generally aligned (rotationally) such that the robots' on board cameras are facing toward back panel 528.
Referring to
In addition, as can be seen in
In addition, as the electrical charging port 300 is being retracted from the electrical charging assembly 200 during the un-mating process, due to the magnetic connection between the electrical charging assembly 200 and the electrical charging port 300 (described above), electrical charging assembly 200 will be pulled in the direction of arrow 555 until the magnetic force is overcome. Gas spring 552 also ensures that the movement is limited, by providing a force in the direction of arrow 554.
While the electrical charging port 300 (which is the female portion of the connector) is described herein to be mounted on the robot and the electrical charging assembly 200 (which is the male portion of the connector) is described herein as being mounted on the charging station, of course, these components could be reversed. In which case the electrical charging port 300 would be mounted on the charging station and the electrical charging assembly 200 would be mounted on the robot. Moreover, as will be apparent to those skilled in the art, other charger ports and designs may be used in connection with the embodiments described herein.
Referring again to
Continuing to refer to
Robot Docking
The docking of a robot to the electrical charging station 500 for recharging, according to one embodiment, is described with regard to
One such precision docking process utilizes the orientation of surfaces 538 and 540 (and fiducials 546 and 548, respectively) relative to cameras 24a and 24 is described with regard to
The optical axis 612 (i.e. the centerline of the field of view or Φ/2) of camera 24b intersects surface 40 and fiducial 48 at a perpendicular angle. In order to ensure that when docked the optical axes of the cameras will be aligned perpendicular to surfaces, 538 and 540, the angle θ which is the orientation of surfaces 538 and 540 relative to surface 536 must be properly set. In this example, the angle θ is approximately 150 degrees. By positioning the fiducials in this manner, the visibility of the fiducials by the cameras 24a and 24b is increased.
As described above, since the cameras are offset from the center of the robot they combine to provide a wide field of view. However, the orientation of the cameras make viewing the fiducials on the charging station challenging. To address this issue, the fiducials may be oriented at an angle to better align with the cameras, which makes the fiducials easier to more accurately read. This may be accomplished by orienting the optical axis of the camera to be at a substantially perpendicular angle to and centered on the fiducial when the robot is in the docked position, as is shown in
Once at pose 600,
(1) Each camera will detect one fiducial: the left and right cameras will detect the left and right fiducials, respectively. The fiducials, once detected, can be transformed internally so that to the robot, they appear to be perfectly perpendicular to the path of the robot (i.e., “flat”, as perceived from the camera, rather than appearing skewed). We can then detect the relative sizes of each fiducial marker, and use that to determine if the robot is closer to one fiducial than the other. This indicates that the robot is not perfectly centered in its approach, and needs to move towards the center line. If we refer to the pixel area of the corrected left fiducial as SL and the pixel area of the corrected right fiducial as SR, then the robot needs to minimize |SR−SL|.
(2) Within the left camera image, the left dock fiducial will be some number of pixels from the right side of the image. We will call this number DL. Likewise, the for the right camera image, the right dock fiducial will be some number of pixels DR from the left side of the image. The robot therefore needs to minimize |DR−DL|.
As the robot needs to correct for the error in (1) first, we issue a constant linear velocity to the robot, and issue a rotational velocity of kS (SR−SL) to the robot until this value gets below some threshold TS. The term kS is a proportional control constant whose value is in the range (0, 1]. When the threshold TS is satisfied, the robot attempts to minimize the error in (2) by issuing a rotational velocity to the robot of kD (DR−DL), where kD is also a proportional control constant in the range of (0, 1]. We continue doing this until either (a) the robot reaches the dock, or (b) the error |SL−SR| grows outside the threshold TS, at which point we switch back to minimizing the error in (1).
The above described precision navigation approach is one example of various approaches that could be used to dock robot 18 with charging station 500. In other embodiments, the precision navigation approach that causes the robot to dock to the electrical charging system may employ techniques similar to those used by the robot more generally when navigating about the warehouse.
The following description of the robot system and robot navigation, including the examples given for navigating the robot to the charging system, is not limiting to the techniques shown and described below for localizing and controlling the robot during precision docking. That is, other techniques for navigating the robot to the initial pose of the charging system may be employed by robots having alternative systems and operation without loss of application of the invention herein to the techniques described for precision docking.
Robot System
Data processor 620, processing modules 640 and sensor support modules 660 are capable of communicating with any of the components, devices or modules herein shown or described for robot system 614. A transceiver module 670 may be included to transmit and receive data. Transceiver module 670 may transmit and receive data and information to and from a supervisor system or to and from one or other robots. Transmitting and receiving data may include map data, path data, search data, sensor data, location and orientation data, velocity data, and processing module instructions or code, robot parameter and environment settings, and other data necessary to the operation of robot system 614.
In some embodiments, range sensor module 662 may comprise one or more of a scanning laser, radar, laser range finder, range finder, ultrasonic obstacle detector, a stereo vision system, a monocular vision system, a camera, and an imaging unit. Range sensor module 662 may scan an environment around the robot to determine a location of one or more obstacles with respect to the robot. In a preferred embodiment, drive train/wheel encoders 664 comprises one or more sensors for encoding wheel position and an actuator for controlling the position of one or more wheels (e.g., ground engaging wheels). Robot system 614 may also include a ground speed sensor comprising a speedometer or radar-based sensor or a rotational velocity sensor. The rotational velocity sensor may comprise the combination of an accelerometer and an integrator. The rotational velocity sensor may provide an observed rotational velocity for the data processor 620, or any module thereof.
In some embodiments, sensor support modules 660 may provide translational data, position data, rotation data, level data, inertial data, and heading data, including historical data of instantaneous measures of velocity, translation, position, rotation, level, heading, and inertial data over time. The translational or rotational velocity may be detected with reference to one or more fixed reference points or stationary objects in the robot environment. Translational velocity may be expressed as an absolute speed in a direction or as a first derivative of robot position versus time. Rotational velocity may be expressed as a speed in angular units or as the first derivative of the angular position versus time. Translational and rotational velocity may be expressed with respect to an origin 0,0 (e.g.
In some embodiments, robot system 614 may include a GPS receiver, a GPS receiver with differential correction, or another receiver for determining the location of a robot with respect to satellite or terrestrial beacons that transmit wireless signals. Preferably, in indoor applications such as the warehouse application described above or where satellite reception is unreliable, robot system 614 uses non-GPS sensors as above and techniques described herein to improve localization where no absolute position information is reliably provided by a global or local sensor or system.
In other embodiments, modules not shown in
One skilled in the art would recognize other systems and techniques for robot processing, data storage, sensing, control and propulsion may be employed without loss of applicability of the present invention described herein.
Maps
Navigation by an autonomous or semi-autonomous robot requires some form of spatial model of the robot's environment. Spatial models may be represented by bitmaps, object maps, landmark maps, and other forms of two- and three-dimensional digital representations. A spatial model of a warehouse facility, as shown in
Spatial models in a warehouse facility may also represent target locations such as a shelf or bin marked with a fiducial to which a robot may be directed to pick product or to perform some other task, or to a temporary holding location or to the location of a charging station. For example,
The spatial model most commonly used for robot navigation is a bitmap of an area or facility.
The scale and granularity of map 720 shown in the
As depicted in
In some embodiments the supervisory system may comprise a central server performing supervision of a plurality of robots in a manufacturing warehouse or other facility, or the supervisory system may comprise a distributed supervisory system consisting of one or more servers operating within or without the facility either fully remotely or partially without loss of generality in the application of the methods and systems herein described. The supervisory system may include a server or servers having at least a computer processor and a memory for executing a supervisory system and may further include one or more transceivers for communicating information to one or more robots operating in the warehouse or other facility. Supervisory systems may be hosted on computer servers or may be hosted in the cloud and communicating with the local robots via a local transceiver configured to receive and transmit messages to and from the robots and the supervisory system over wired and/or wireless communications media including over the Internet.
One skilled in the art would recognize that robotic mapping for the purposes of the present invention could be performed using methods known in the art without loss of generality. Further discussion of methods for robotic mapping can be found in Sebastian Thrun, “Robotic Mapping: A Survey”, Carnegie-Mellon University, CMU-CS-02-111, February, 2002, which is incorporated herein by reference.
Scans
A robot outfitted with sensors, as described above, can use its sensors for localization as well as contribute to the building and maintenance of the map of its environment. Sensors used for map building and localization may include light detection and ranging (“LIDAR” or “laser scanning” or “laser-radar”) sensors. Laser-radar scanners measure the range and distance to objects in a horizontal plane with a series of discrete, angular sweeps of the robot's local environment. A range finding sensor acquires a set of measurements, a “scan” taken at discrete angular increments of preferably one-quarter (0.25) degree increments over a 180-degree arc or a greater or lessor degree arc, or a full 360-degree arc about the robot. A laser-radar scan, for example, may be a set of measurements representing the return time and strength of a laser signal, each measurement at a discrete angular increment indicating a potential obstacle at a distance from the robot's current position.
For illustration, as shown in
Other forms of range finding sensors include sonar, radar, and tactile sensor without departing from the scope of the invention. Examples of commercially available range finding and location and orientation sensors suitable for use with the present invention include, but are not limited to, the Hokuyo UST-10LX, the SICK LMS 100, and the Velodyne VLP-16. A robot may have one or more range or location sensors of a particular type, or it may have sensors of different types, the combination of sensor types producing measurements that collectively map its environment. Further discussion of methods of robotic mapping by LIDAR and other scanners can be found in Edwin B. Olson, “Robust and Efficient Robotic Mapping”, PhD Dissertation, Carnegie-Mellon University, 2008, which is incorporated herein by reference.
Scan Matching
“Scan matching” is the process of comparing range finding scans by different robots or scans of a single robot taken at different times or to a map of an environment such as a SLAM map. In the scan-to-scan matching process, a first laser-radar scan taken by a robot at one time may be compared to a second, earlier scan to determine if the robot has returned to the same location in the map. Likewise, matching the scan to a second robot's scan can determine if the two robots have navigated to a common location in the map. Scan matching to a map can be used to determine the pose of the robot in the mapped environment. As illustrated in
It is unlikely that a laser-radar scan matches exactly with the map at any arbitrary location and orientation. Uncertainties in sensor measurements, the demands of pose accuracy, and limited computational cycle times require robust and efficient algorithms to statistically determine the best scan match between a robot's sensed environment and its actual pose. Statistical methods, however, are susceptible to producing inaccurate poses and can be computationally expensive.
Various methods and algorithms have been developed to address these complexities. A survey of scan matching techniques and a two-resolution method for ensuring accuracy while reducing computational complexity in scan matching for localization can be found in Edwin B. Olson, “Real-Time Correlative Scan Matching”, in Proceedings of the 2009 IEEE international Conference on Robotics and Automation (ICRA'09), IEEE Press, Piscataway, N.J., USA, 2009, pp. 1233-1239, which is incorporated herein by reference.
M3RSM
As previously mentioned, another such technique for localizing using scan matching is many-to-many multiresolution scan matching or “M3RSM”. M3RSM extends the two-resolution correlative scan matching approach to multiple resolutions, using a pyramid of maps, each constructed by decimation for computational efficiency. A discussion of M3RSM can be found in Edwin Olson, “M3RSM: Many-to-many multi-resolution scan matching”, Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), June 2015, which is incorporated herein by reference. M3RSM dramatically reduces the processing time to perform scan matching against a SLAM map by eliminating candidate poses from consideration at lower resolutions of the map. Robot localization and navigation along a goal path to a goal pose is further improved by minimizing the M3SRM search space according to methods disclosed in related U.S. application Ser. No. 15/712,222, entitled MULTI-RESOLUTION SCAN MATCHING WITH EXCLUSION ZONES, filed on Sep. 22, 2017, and incorporated by reference in its entirety herein.
Navigating to a Charging Station
As shown and described with reference to
During navigation of the robot to the charging station, the robot may navigate to location 708 as it would for any other target pose associated with a target product bin or fiducial as above described.
Continuing with navigation process 800, at step 806 the robot, using robot system 614, receives a goal pose, for example the pose 600 of a charging station 500 at location 718. At step 808, robot system 614 then generates, using path planning module 646, the goal path from its initial pose to the pose associated with the charging station. The goal path may then be stored for later processing. In some embodiments, the goal path may be generated based on a pose estimate for the robot, or, preferably, generating the goal path is based on the pose of the robot determined after the first iteration of the “Find Pose” step 812. Path planning module 642 may generate the goal path from the current pose to the goal pose by a variety of techniques known to practitioners in the art including the A* and D* pathfinding algorithms. Alternatively, the robot may receive a goal path via transceiver module 670 or may retrieve a goal path from data storage 630. Having received the map and generated the map pyramids and goal path, robot system 614 may then proceed to move the robot incrementally along the goal path.
At step 810, the robot receives a laser-radar scan of the local environment and proceeds to find the pose best matching the received scan. As illustrated above with reference to
At step 812, “Find Pose”, the current pose of the robot is found. First, a search area is determined as the portion of the received map to be searched for candidate poses. In a first iteration, the search area may include the entire map. In a subsequent iteration, the robot may estimate its pose within only a portion of the map. The pose for determining the search area may be estimated from a last known pose combined with sensor data such as drive train/wheel encoders and/or drive control information. One skilled in the art would understand that estimates of pose and determining the search area could be performed by various methods and parameters. In a preferred embodiment, state estimation module 650 may fuse pose data with wheel encoder data and inertial sensor data to determine the robot's current pose, velocity, and estimated errors for each. The estimated pose thus bounds the search to a portion of the map, reducing the search space and decreases processing time for scan matching. The lower the uncertainty in the pose estimate, the smaller the search area over which scan matching may be required. The larger the uncertainty, the greater the search area over which scan matching may be required. Next, the pose within the search area is determined by scan matching according to scan matching techniques such as M3RSM as referenced above. At optional step 814, the process may return to step 808 to generate or update the goal path based on a first or subsequent find pose result.
Having found the current pose of the robot, continuing to step 816 of
Where the goal path includes at a goal pose that is the pose assigned to a charging station, the process may continue with precision docking as follows.
Precision Docking with High Resolution Localization
Mating the electrical charging assembly and electrical charging port, according to the above-disclosed dimensions of one embodiment of the electrical charging assembly and electrical charging port, may require higher resolution maps than the maps used for warehouse navigation. That is, the navigation approach used by the robot to arrive at pose 600, which may use 5 cm-resolution maps, for example, may not precisely position the robot at mating pose 602, such that the electrical charging assembly 200 of charging station 500 and the electrical charging port 300 of robot 18 are reliably mated. Using the 5-cm resolution for localization and scan matching may also require that the charging station be perfectly mapped and firmly fixed to the warehouse floor.
Thus, in an embodiment of precision docking, upon arriving at pose 600 of charging station 500 the robot 18 may switch to using a higher resolution SLAM map of the environment, preferably a 1 cm-resolution SLAM map, and localizing by scan matching techniques as described above. Localization using a higher resolution map, such as a 1 cm-resolution map, may proceed as described with reference to process 830 of
At step 836,
While providing for precision localization when docking to the charging station, using a higher resolution map adds computational complexity and robot system data processor and data memory resource demands. For example, the processing demands for localizing by scan matching on a 1 cm-resolution map demands as much as 25-times the computation of using a 5 cm-resolution map. Thus, making use of a higher resolution map for localization by scan matching during docking wastes processing time that could be used for other critical processing tasks. Furthermore, in the area of the charging station, the map of the entire warehouse is not needed once the robot is proximate to the charging station. Still more, navigation by scan matching to the entire warehouse map, assuming it includes a map of the charger docking station, would not be tolerant of movement of the charging station during docking.
Accordingly,
For example, as shown in
It is noted that the charging station, in other embodiments, may be in other dimensions and configurations, i.e. the side surfaces at the back of the docking station may not be angled relative to the center surface. Indeed, the methods described herein for docking a robot to a charging station may be applied to other dimensions and configurations of a charger docking station without loss of generality. With such other dimensions and configurations, the map of the charging station used for scan matching need only provide a scan map including or consisting solely of a scan or scan representation of a charger docking station that matches a range-finding scan of the robot. Such robots may use other range finding scanning methods consistent with producing a local scan for scan matching to the map of the charging station.
In view of the navigation process 830 described above with reference to
In one embodiment, obstacle avoidance during docking, when navigating from the initial pose to the mating pose, may be simplified by determining from each local scan, at each iteration, whether an obstacle exists within a threshold distance d, where d is less than the distance that the robot can get to the charger docking station when fully mated. An obstacle appearing in the local scan within distance d is thus not the docking station itself. For example, as shown in
By localizing against the charging station only, the robot may perform precision docking at maximum efficiency for the short duration of the final approach to the charger docking station. Localizing against the charging station only may be used in conjunction with higher resolution maps while docking, and may be used with other robot control techniques, such as “arc control” to be further described below, without loss of generality as to the inventive aspects of “localizing to the dock.”
Precision Docking with Arc Control
Precision docking according to the embodiments described above with reference to
Unfortunately, there are an infinite number of circles with radius r having an arc section passing through XR, YR and XD, YD. By introducing the constraint that the tangent to the circle at pose XD, YD must have a slope of tan(θD), i.e., the robot's final orientation is perpendicular to the charging station, and further utilizing the constraint that the center XC, YC of circle 764 will be the same distance from XR, YR and XD, YD, radius r can be found as follows:
A third constraint provides that the equation of the line passing through XD, YD and XC, YC has a slope that is perpendicular to the tangent line slope of tan(θD). Defining variable p as follows:
and solving for XC and YC:
provides for solving for radius r by simple substitution into equations (1) or (2) above.
As above, the radius r of the unique circle having center XC, YC passing through XR, YR and XD, YD defines the desired arc path 762 from pose 604 to mating pose 602. Thus, the control for an incremental movement of the robot along path 762 may be determined from the tangent line of the circle 764 at each iteration. That is, the control of the robot at each iteration may be found by advancing the robot in the direction of the tangent line at an instantaneous location x′R, y′R, at an angular velocity θ′T, where θ′T is the tangent to the circle 764 at x′R, y′R.
In practice, some variation in the actual path may occur as the robot moves incrementally from pose 604 to pose 602 along the control path 762. The instantaneous velocity of the robot at each incremental pose along path 762 should, however, result in an instantaneous trajectory within a small error from the control path. For example,
To ensure that radius r does not change, and observing that:
where x′R is the instantaneous linear velocity of the robot and θ′R is its instantaneous angular velocity, for a given radius r, the instantaneous linear velocity x′R may be held fixed by adjusting instantaneous angular velocity θ′R, or angular velocity θ′R may be held fixed by adjusting linear velocity x′R. Thus, by issuing a control to the robot according to:
θ′R=kΦ (8)
where k is a proportional control constant, and combining the rotational controls from equations (7) and (8) above:
where α and β are weighting parameters, the combined control equation (9) closes the error between the robot's actual path 772 and the desired arc path 762. In a preferred embodiment, the weighting parameters α and β may be one (1).
As robot 18 gets nearer to the charging station, the proportional control of equation (8) may be accounted for more heavily in equation (9). In another embodiment, weighting parameters α and β may be adjusted in nonlinear relation as a function of the distance to the charging station. Alternatively, the control scheme may be applied by first closing the rotational error according to equation (8) until the error gets below a threshold, then setting X′R to a fixed value, and next controlling the robot according to equation (7), constantly updating r and Φ, and then switching the control scheme back to equation (8) when the threshold is again exceeded. In this manner, the error Φ in the trajectory of the robot along arc path 762 and at final pose 602 is minimized.
While the foregoing description of the invention enables one of ordinary skill to make and use what is considered presently to be the best mode thereof, those of ordinary skill will understand and appreciate the existence of variations, combinations, and equivalents of the specific embodiments and examples herein. The above-described embodiments of the present invention are intended to be examples only. Alterations, modifications and variations may be effected to the particular embodiments by those of skill in the art without departing from the scope of the invention, which is defined solely by the claims appended hereto.
It should be understood that the present invention may be implemented with software and/or hardware. Accordingly, aspects of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects that may all generally be referred to herein as a “circuit,” “module” or “system.” As will be appreciated by one skilled in the art, aspects of the invention may be embodied as a system, method or computer program product.
Aspects of the present invention are described with reference to flowcharts, illustrations and/or block diagrams of methods and apparatus (systems). The flowcharts and block diagrams may illustrate system architecture, functionality, or operations according to various embodiments of the invention. Each step in the flowchart may represent a module, which comprises one or more executable instructions for implementing the specified function(s). In some implementations, steps shown in succession may in fact be executed substantially concurrently. Steps may be performed by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
Computer instructions for execution by a processor carrying out operations of the present invention may be written one or more programming languages, including object-oriented programming languages such as C#, C++, Python, or Java programming languages. Computer program instructions may be stored on a computer readable medium that can direct the robot system via the data processor to function in a particular manner, including executing instructions which implement the steps specified in a flowchart and/or system block diagram described herein. A computer readable storage medium may be any tangible medium that can contain, or store instructions for use by or in connection with the data processor. A computer readable medium may also include a propagated data signal with computer readable program code embodied therein.
The invention is therefore not limited by the above described embodiments and examples, embodiments, and applications within the scope and spirit of the invention claimed as follows.
This application is related to pending U.S. application Ser. No. 15/712,491, filed Sep. 22, 2017, entitled “AUTONOMOUS ROBOT CHARGING STATION”, which is incorporated herein by reference. This application is related to co-filed U.S. application Ser. No. 15/821,669 filed Sep. 22, 2017, entitled “ROBOT CHARGER DOCKING LOCALIZATION”, which is incorporated herein by reference.
