Patent Application
20210387341
The present disclosure relates to measures for avoiding collisions involving a modular robot. In particular, the present disclosure relates to automatically avoiding collisions involving a modular robot that has been reconfigured.
US 2017/0348856 A1 teaches a collision avoidance procedure involving a geometric model. When the geometric model is determined to interfere with a no-entry area, a repulsive force for repelling an interfering portion of the geometric model from an interfering portion of the no-entry area is virtually generated, a posture in a state where the interfering portion of the model is pushed from the no-entry area into the operating area by the repulsive force is calculated, and the calculated posture is used as a way-point.
The present invention provides an anti-collision safety device and methods for a modular robot. The anti-collision safety device and the methods may automatically derive a new/updated geometric model of the robot whenever a module is removed or added to the robot.
The anti-collision safety device is configured to assign modules of the robot to elements of a kinematic chain, and to determine a geometric model for the elements of the kinematic chain based on data describing the geometry of the assigned modules. The device may then calculate a collision-free movement of elements of the kinematic chain based on the determined geometric model, a model of the environment, and a target position and/or orientation of at least one element of the kinematic chain.
These actions of the anti-collision safety device may be triggered by an indication that the robot has been reconfigured, i.e., that a module (e.g., a joint module or a link module) has been removed or added to the robot.
In this regard, the term “joint module”, as used throughout the description and the claims, particularly refers to a chain of two or more rigid bodies, which are connected to each other by one or more actuatable joints. An actuatable joint allows controlling a position of a rigid body at a distal end of the actuatable joint relative to another rigid body at a proximal end of the actuatable joint. A motion of the distal end of a joint module relative to a proximal end of the joint module may be restrained to one, two or three (or even more) degrees of freedom. Moreover, the term “link module”, as used throughout the description and the claims, particularly refers to a single rigid body. The joint and link modules both comprise a proximal end and a distal end, wherein at least the proximal end is attachable to another module. Hence, the terms “modular robot”, or “robot”, as used throughout the description and the claims, particularly refer to an assembly of joint modules and optionally one or more link modules connected at their proximal/distal ends. The modules may be regarded as a kinematic chain of rigid bodies connected by actuatable joints which provide constrained (or desired) motion.
Furthermore, the term “geometric model”, as used throughout the description and the claims, particularly refers to a chain of rigid bodies (shapes) which may be defined by one or more geometric primitives arranged relative to each other within a common coordinate frame. In addition, the formulation “collision-free movement of elements of the kinematic chain”, as used throughout the description and the claims, particularly refers to avoiding unwanted (hazardous) collisions, as opposed to desired object manipulations performed by the robot. Moreover, the term “element of a kinematic chain”, as used throughout the description and the claims, particularly refers to a rigid body representing one or more modules of the robot.
By (automatically) deriving a new/updated geometric model for the kinematic chain when a module (e.g., a joint module or a link module) has been removed or added to the robot, a collision risk can be reduced, and the robot safety can be increased.
The anti-collision safety device may further be configured to retrieve the data describing the geometry of the assigned modules from the assigned modules. The data may define different shapes (virtual rigid bodies) for one or more of the modules (i.e., each module may have at least one shape assigned thereto). The different shapes may differ in regard to fidelity (e.g., high, low, etc.).
For example, each module may be provided with a persistent memory and data on the geometry of the module may be stored in said memory. In addition, data on shapes approximating the geometry may be stored in said memory. The shapes may differ in size and attention to detail, wherein a smaller shape may comprise more geometric details than a larger shape approximating said geometric details.
The anti-collision safety device may further be configured to extract transformations for the assigned modules from the data. Each of said transformations may correspond to one module and relate a proximal end of said module to a distal end of said module.
The anti-collision safety device may be further configured to determine the geometric model of the kinematic chain based on the extracted transformations.
The transformations allow determining the position of each module inside the robot and thus enable automatically generating the geometric model. In addition, the transformations can be used to determine the kinematic model (through generating the Denavit-Hartenberg parameters) and/or the dynamic model. Using the transformations, the geometric model can be generated independently of the complexity of the geometry of each module.
The anti-collision safety device may further be configured to update said data stored in the modules through a wired or a radio interface.
For example, a shape approximating the geometry may be replaced with another shape or a new shape may be added to the shapes stored in the memory.
The anti-collision safety device may further be configured to plan, for each of the modules of the robot, a first part of an intended path in a dynamic environment, the first part of the intended path avoiding regions in which collisions might occur. The anti-collision safety device may further be configured to plan, for each of the modules of the robot, a fail-safe maneuver at the end of the first part of the intended path, the fail-safe maneuver avoiding regions in which collisions might occur.
Collisions might occur with both static and dynamic obstacles. Notably, the trajectory of dynamic obstacles may be predicted.
The anti-collision safety device may further be configured to plan, for each of the modules of the robot, a second part of the intended path in the dynamic environment. The anti-collision safety device may further be configured to execute the fail-safe maneuver for at least those modules of the robot for which it cannot be ascertained, before the end of the first part is reached, that the second part avoids regions in which collisions might occur.
The fail-safe maneuver may be directed at causing a standstill of the robot at or before an end of the planned fail-safe maneuver.
The anti-collision safety device may further be configured to abort the fail-safe maneuver if a recovery maneuver becomes available which avoids regions in which collisions might occur.
The recovery maneuver may bring a module back onto the intended path or at least closer to the intended path, compared to following the fail-safe maneuver.
The safety device may further be configured to determine the geometric model based on approximating the geometry of the modules when being provided with said data.
I.e., the safety device may (automatically) reduce the complexity of the geometric model to decrease the computational load.
The safety device may further be configured to identify a module that has been added to the modular robot and to access a database that stores data describing the geometry of said module.
For example, a joint module and/or a link module may be provided with a memory element that stores a manufacturer and part identification (ID) which enables the safety device to search and retrieve data describing the geometry of said module.
For example, a manufacturer may provide a database accessible via the internet and may regularly update the database if new joint modules/link modules become available. In another example, the data may be stored within the memory element, thereby allowing the safety device to (automatically) retrieve the data from the joint module or the link module. Notably, storing the data within the joint module or the link module reduces the risk of data tampering and hence further increases the overall safety.
The safety device may further be configured to identify a module that has been added to the modular robot by receiving a control signal via a wired or a wireless connection from said module.
For example, an added module may integrate itself into a communication network and transmit the control signal to the device via the network. For instance, the safety device may act as a master device that polls data from the network elements, or the element may broadcast the signal via the network (e.g., as part of an initialization or startup routine).
The safety device may be configured to identify the module that has been added to the modular robot by receiving the control signal via a bus interface from said module.
For instance, the modules of the modular robot may form a daisy chain or a logical ring network and a control frame (i.e., a control message with a header identifying the control data in the body) from the safety device may be circulated (from module to module) along the chain with each element adding identification data to the control frame such that the safety device learns the structure of the robot by evaluating the control frame. In another example, modules may identify themselves and their neighbors.
The safety device may further be configured to identify a module that has been added to the modular robot by evaluating a signal from a sensor of the safety device.
For example, the safety device may be provided with an RFID sensor that scans the vicinity for RFID tags attached to the modules. Hence, whenever a new RFID tag is detected, the safety device may assume that a new module (as identified by the RFID tag) has been added to the modular robot. To further increase safety, the (growing) kinematic chain may be depicted to a user and the user may take corrective action, if necessary.
The safety device may be configured to identify the module that has been added to the modular robot by analyzing an image of at least a part of the module or by activating a radio-frequency identification, RFID, tag reader.
For instance, the modules may be provided with a bar code, a matrix code, letters, numbers (or other graphical codes) that convey information which allows identifying the modules. In another example, the safety device may identify the module based on matching its shape to one of a plurality of possible shapes.
The safety device may further be configured to identify a module that has been added to the modular robot by receiving a manual user input.
For example, the user may be provided with a user terminal (connected to the safety device) that allows assembling/reconfiguring a virtual modular robot, wherein a layout of the virtual modular robot is to reflect (changes to) the structure of the physical robot. Notably, the safety device may be part of or integrated into a control system/apparatus which controls the operation (and particularly the motion) of the robot.
A first method comprises reconfiguring the robot by removing one or more of first link and/or joint modules from the modular robot and adding one or more second link and/or joint modules to the modular robot, determining a geometric model of the robot based on data describing the geometry of the one or more second link and/or joint modules and while operating the robot, using the geometric model to plan collision-free paths of the remaining first and one or more second link and/or joint modules.
The data may comprise a series of transformations corresponding to the remaining first and one or more second link and/or joint modules and a pose of one of the remaining first and one or more second link and/or joint modules in a reference frame may be given by applying the transformations of the series up to a transformation corresponding to the one of the remaining first and one or more second link and/or joint modules.
Determining the geometric model of the robot may comprise determining positions of the one or more second link and/or joint modules in a kinematic chain represented by the geometric model.
The remaining first and one or more second link and/or joint modules may comprise nodes of a wired communication network.
Determining the position of the one or more second link and/or joint modules in the kinematic chain may comprises identifying one or more nodes neighboring a node comprised in the one or more second link and/or joint modules.
Determining the positions of the one or more second link and/or joint modules within the kinematic chain may comprise taking one or more images of the robot and determining a type of the remaining first and one or more second link and joint modules detected in said one or more images.
Determining the positions of the one or more second link and/or joint modules within the kinematic chain may comprise scanning the robot for radio-frequency identification, RFID, tags attached to or embedded into the one or more second link and/or joint modules.
The first method may further comprise collecting said data from the remaining first and one or more second link and/or joint modules and/or updating said data through a wired or a wireless connection.
Determining the geometric model of the robot may comprise approximating two or more neighboring link and/or joint modules of the robot as a single virtual body representing an element of the kinematic chain.
Approximating two or more neighboring link and/or joint modules of the robot as a single virtual body may involve selecting between different approximation levels.
The geometric model may comprise a chain of virtual bodies, of which each virtual body represents an element of the kinematic chain.
The first method may further comprise determining a plurality of link and joint configurations from the first and second link and joint modules, determining, for each of said configurations, whether said configuration enables the robot to perform a set of subtasks of a task, determining, for each configuration that enables the robot to perform the set of subtasks, whether said configuration enables the robot to perform the task, determining, for each configuration that enables the robot to perform the task, a value indicating a cost of performing the task, according to a cost metric, by carrying out a simulation based on a/the geometric model and using the determined values to select a configuration that enables the robot to perform the task.
The first method may further comprise assigning a category to the task which is to be performed by the robot, wherein the category is selected from the group consisting of tasks with given tool paths, tasks with partially given tool paths and tasks with yet to be defined tool paths within a given workspace, wherein determining whether said configuration enables the robot to perform the task comprises determining, whether said configuration enables the given tool path, the partially given tool path and any pose within the workspace, respectively.
A second method comprises reconfiguring the robot by adding and/or removing link modules and/or joint modules to/from the robot, determining a geometric model for the robot based on data describing the geometry of the modules mounted to a frame after said reconfiguration, and using the geometric model, when the robot is operated, for collision avoidance.
In this regard, the term “frame”, as used throughout the description and the claims, particularly refers to a part of the robot that remains static throughout operation and which may contain the control logic that controls the actuation of the joints. As for a mobile robot, the term “frame”, as used throughout the description and the claims, particularly refers to a part of the robot to which the elements of the kinematic chain are attached, thereby forming one or more limbs of the robot which extend from the frame.
Determining the geometric model for the robot may comprise identifying a layout of the robot and accessing a database for retrieving said data.
As described above, a link module or a joint module may be provided with a memory element that holds a manufacturer and part identification (ID) which allows the safety device to search and retrieve data describing the geometry of said element. For example, a manufacturer may provide a database accessible via the internet and may regularly update the database if new link modules/joint modules become available. In another example, the data may be stored within the memory element, allowing the safety device to (automatically) retrieve the data from the link module or the joint module.
Determining the geometric model for the robot may comprise approximating one or more elements of the kinematic chain that covers two or more neighboring modules of the robot.
For example, neighboring modules that cannot or will not be moved relative to each other may be modeled (approximated) by a single virtual body representing the corresponding element of the kinematic chain. In another example, neighboring modules that may be moved relative to each other may be modeled (approximated) by a single virtual body representing an element of the kinematic chain that covers all movements that may occur.
Identifying the layout may comprise identifying a position of a link module or a joint module within a kinematic chain.
The link modules and/or the joint modules may form a daisy chain or logical ring network and identifying the position of the link module or the joint module within the kinematic chain may comprise identifying a position of the link module or the joint module within the network.
As described above, a control frame from the safety device may be circulated (from module to module) along the modular robot with each module adding identification data to the control frame such that the safety device learns the structure of the robot by evaluating the control frame. In another example, elements may identify themselves and their neighbors.
In yet another example, identifying the position of the link module or the joint module within kinematic chain may comprise taking images of the modules of the robot.
As described above, the modules may be provided with a bar code, a matrix code, letters and/or numbers that convey information which allows identifying the modules. In another example, the safety device may identify the module based on matching its shape to one of a plurality of possible shapes.
Using the geometric model, when the robot is operated, for collision avoidance may comprise calculating a collision free path for moving one or more modules based on the geometric model, wherein the geometric model comprises a chain of approximated virtual rigid bodies that can be displaced relative to each other.
It will be appreciated that the features and attendant advantages of the disclosed methods may be realized by the disclosed device and vice versa. Moreover, it is noted that throughout the description, features in brackets are to be regarded as optional.
The foregoing aspects and many of the attendant advantages will become more readily appreciated as the same becomes better understood by reference to the following detailed description, when taken in conjunction with the accompanying drawings, wherein like reference numerals refer to like parts throughout the various views, unless otherwise specified.
Notably, the drawings are not drawn to scale and unless otherwise indicated, they are merely intended to conceptually illustrate the structures and procedures described herein.
At step 4, the process continues with determining, for each configuration that enables the robot 10 perform the set of subtasks, whether said configuration enables the robot 10 to perform the task. I.e., those configurations that have passed the previous test may be further tested to see whether all subtasks can be performed (for example, whether all required positions can be reached). At step 6, the process is continued by determining, for each configuration that enables the robot 10 to perform the task, a value indicating a cost of performing the task, according to a cost metric. This may be achieved by carrying-out a simulation based on a geometric model representing the configuration. At step 8, the determined values are used to select a configuration that enables the robot 10 to perform the task.
The distal end of the second link module 16b is connected to a second joint module 18b (hinge joint) which allows for swiveling a third link module 16c attached to the second link module 16b by the second joint module 18b around a second horizontal axis (not shown). The third link module 16c is connected to a tool 20 (e.g., a claw or another manipulator). When the tool 20 is to be moved to a target position, the anti-collision safety device 12 ensures that none of the modules 16a-16c, 18a, 18b or the tool 20 collides with obstacles 22 in the surrounding. To reduce the computational complexity of this task, the anti-collision safety device 12 may overapproximate the geometry of the modules 16a-16c, 18a, 18b and the tool 20 (as illustrated in the right part of
The anti-collision safety device 12 may only allow a movement of the modules 16a-16c, 18a, 18b and the tool 20, if a corresponding movement of the first and second virtual rigid bodies 24a, 24b does not cause a collision with an obstacle 22. For example, movements (or a sequence of movements) of the modules 16a-16c, 18a, 18b and the tool 20 may be validated based on simulating corresponding movements of the virtual rigid bodies 24a, 24b of the (simplified) geometric model 26 within (or as part of) a model of the environment. For example, the anti-collision safety device 12 may be provided with a mono- or stereo-camera system, a laser-scanner, a lidar, an ultrasonic distance device, etc. that recognizes/tracks (static and/or dynamic) obstacles (including humans) 22 in the range of the robot 10.
When the anti-collision safety device 12 receives an indication that the robot 10 has been reconfigured, it may check whether the geometric model 26 must be updated. For example, as illustrated in the lower part of
By assigning the link modules 16a-16d and the joint modules 18a-18c of the robot 10 to elements 24a, 24b of a kinematic chain on which the geometric model 26 is based, the anti-collision safety device 12 may assess the structural change and determine a new/updated geometric model 28 for the elements of the kinematic chain. Notably, the indication that the robot 10 has been reconfigured may be caused by a self-check of the robot 10 or may be manually input, as will be explained in more detailed with reference to
For example, as illustrated in
That is, the anti-collision safety device 12 may transmit an empty control frame to the electronic circuit 30a mounted to the first joint module 18a. Upon receiving the control frame, the electronic circuit 30a may add its ID and/or geometric data describing the shape of the first joint module 18a to the control frame and transmit the (modified) control frame to the electronic circuit 30b in the neighboring link module 16d, that acts likewise. When the control frame has been passed through the electronic circuits 30a-30c, it is received by the electronic circuit 30d of the tool 20 (which is the last one of the electronic circuits 30a-30d in the chain).
As the electronic circuit 30d has no neighboring element at its distal end, it may append its ID and/or geometric data describing the shape of the tool 20 and transmits the (thus modified) control frame along the chain of electronic circuits 30a-30d back to the anti-collision safety device 12. The anti-collision safety device 12 may then derive the layout of the robot 10 by analyzing the sequence of entries in the control frame. In another example, only the link modules 16a-16d or only the joint modules 18a-18c may be provided with electronic circuits, if, for example, they can provide data on the joint modules 18a-18c and the link modules 16a-16d, respectively, to which they are connected. In other words, all elements 24a, 24b of the kinematic chain should be covered by at least one electronic circuit, but one electronic circuit may cover more than one element 24a, 24b.
Furthermore, instead of transmitting a control frame along a daisy chain or a logical ring, each electronic circuit 30a-30d may transmit its ID and/or geometric data describing the module shape and/or information on the modules with which it is connected to the anti-collision safety device 12. For example, the electronic circuits 30a-30d may be connected to a bus or transmit the data wirelessly. In another example shown in
In another example shown in
The anti-collision safety device 12 may also extract transformations for the assigned modules 16d, 18a, and 18b from the data. Each transformation may correspond to one of the modules 16d, 18a and 18b and relate a proximal end of said module 16d, 18a or 18b to a distal end of said module 16d, 18a or 18b. The transformations may be used to automatically compute the geometric model. Moreover, the anti-collision safety device 12 may update the data stored in the memories 17 through the radio interface 12.
As the module 16d, 18a and 18b travels along the first part 36a the anti-collision safety device 12 may plan a second part 36b of the intended path and a fail-safe maneuver 38b at the end of the second part 36b. The anti-collision safety device 12 may execute the fail-safe maneuver 38a if cannot be ascertained, before the end of the first part 36a is reached at T2, that the second part 36b avoids regions 42 in which collisions might occur, as shown in in the bottom part of
| Number | Date | Country | Kind |
|---|---|---|---|
| 18201371.4 | Oct 2018 | EP | regional |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2019/078463 | 10/18/2019 | WO | 00 |
