A computer-implemented method for operating multiple mobile agents in a coordinated manner

Abstract

The invention proposes a computer-implemented method for operating a plurality of mobile agents (26) in an environment, comprising: a) generating mission information that is indicative of any of a number of mobile agents (26), a pickup location, a drop off location, and at least one preset-time, wherein each mobile agent (26) is associated with a depot, wherein the preset-time is indicative of the maximum time period in which the number of mobile agents (26) has to complete a movement phase; b) at least one movement phase including, for each of the number of mobile agents (26), generating a control command (24) based on the mission information (14), wherein the control command (24) causes the respective mobile agent (26) to move from a starting location (16) to a destination in accordance with the mission information (14); wherein each control command (24) is generated such that the respective mobile agent (26) arrives at its destination within the preset-time.

Description

[0001] DESCRIPTION

[0002] A computer-implemented method for operating multiple mobile agents in a coordinated manner

[0003] TECHNICAL FIELD

[0004] The invention relates to a computer-implemented method for operating a plurality of mobile agents in an environment, in order to coordinate their movement and task performance.

[0005] BACKGROUND

[0006] US 9 469 476 B1 discloses a method for dynamically configuring a delivery system for delivering products to smart mats. A delivery coordination server determines a location of multiple smart mats. Each of the smart mats includes a positioning system and a transmitter that transmits a message describing the real-time geophysical location of the smart mats. The delivery coordination server receives a message describing a location of a first delivery vehicle that is transporting a first package addressed for delivery to a first smart mat. The delivery coordination server determines that the first smart mat has moved to a location that is within a predetermined distance of a second smart mat, to which a second delivery vehicle is scheduled to deliver a second package. The delivery coordination server directs the first delivery vehicle to transfer the first package to the second delivery vehicle for delivery to the first smart mat.

[0007] US 11 157 866 B2 discloses a method for intelligent package delivery. The method comprises obtaining order information for a plurality of package delivery orders, wherein the order information includes one or more package pickup stops and one or more package drop off stops; and determining one or more optimized package delivery route for one more vehicle using the order information, wherein an optimized package delivery route of the one or more optimized package delivery route includes a first stop associated to a first order of the plurality of package delivery orders and a second stop associated to a second order of the plurality of package delivery orders. The known approaches require the individual mobile agents to be individually capable of performing their respective tasks, such as package delivery. There are no time window restrictions in the known approaches.

[0008] SUMMARY OF THE INVENTION

[0009] It is the object of the invention to improve functionality, preferably coordination, of multi-agent systems.

[0010] The invention provides a computer-implemented method for operating a plurality of mobile agents in an environment, the environment comprising a set of depots, a set of pickup locations and a set of drop off locations, wherein the method comprises: a) generating mission information that is indicative of any of a number of mobile agents, a pickup location, a drop off location, and at least one preset-time, wherein each mobile agent is associated with a depot, wherein the preset-time is indicative of the maximum time period in which the number of mobile agents has to cooperatively complete a movement phase; b) at least one movement phase including, for each of the number of mobile agents, generating a control command based on the mission information, wherein the control command causes the respective mobile agent to move from a starting location to a destination in accordance with the mission information; wherein each control command is generated such that the respective mobile agent arrives at its destination within the preset-time.

[0011] With this approach it is possible to coordinate multiple mobile agents, such as mobile robots or vehicles, to collectively perform a task that could not be performed by a single mobile agent alone, such that the interacting mobile agents cooperatively complete task(s) within a preset-time. For example, multiple mobile robots could be needed to have enough weight capacity to carry a certain package. In case of passengers, it may be necessary to pick up more than a single vehicle can carry. The preset-time and the control commands allow for multiple mobile agents to arrive more or less simultaneously at their starting location and destination. It is possible to simultaneously form multiple groups of mobile agents that are going to perform their tasks. As a result, the overall capacity and functionality of a system having multiple mobile agents can be improved.

[0012] Preferably, in step b) the control command is generated such that, after the collective arrival of the mobile agents at the destination, the mobile agents perform a task. The task performance may be automated or contingent on human interaction. For example, the mobile agents may pick up a package. In case of passengers, the mobile agents may wait for the passengers to board the vehicle.

[0013] Preferably, in step b) a first movement phase called an active phase includes, for each of the number of mobile agents, generating a control command based on the mission information, wherein the control command causes the respective mobile agent to move from its depot as the starting location to a pickup location as its destination in accordance with the mission information, wherein each control command is generated such that the respective mobile agent arrives at the pickup location within the preset-time. With this approach delays in the system can be reduced, since the mobile agents can arrive (nearly) simultaneously at the pickup location in a coordinated manner in order to be able to pick up the package or passengers. It should be noted that in general, as disclosed herein, a single mobile agent is insufficient to perform the required task.

[0014] Preferably, in step b) a second movement phase called a handshaking phase includes, for each of the number of mobile agents, generating a control command based on the mission information, wherein the control command causes the respective mobile agent to move from a pickup location to a drop off location in accordance with the mission information, wherein each control command is generated such that the respective mobile agent arrives at the pickup location within the preset-time. With this approach the actual task is complete within a given timewindow and reduced delays, since the resources provided by the mobile agents can be used in a more time efficient manner.

[0015] Preferably, in step b) a third movement phase called a bye-bye phase includes, for each of the number of mobile agents, generating a control command based on the mission information, wherein the control command causes the respective mobile agent to move from a drop off location back to its depot, wherein each control command is generated such that the respective mobile agent arrives at its depot within the preset-time. With this approach a more efficient flow of packages or passengers can be achieved since as soon as the task is completed, the mobile agents clear the environment and get ready for the next mission.

[0016] Preferably, the control command is generated such that, after the collective arrival of the mobile agents at the pickup location, the mobile agents collectively pick up a package or a passenger. Preferably, the control command is generated such that, after the collective arrival of the mobile agents at the pickup location, the mobile agents collectively transport a package or a passenger to the drop off location.

[0017] Preferably, the control command is generated such that, after the collective arrival of the mobile agents at the drop off location, the mobile agents collectively drop off a package or a passenger and the mobile agents return to their respective depots. Coordinated movement, pickup and drop off procedures as well as return to the depots improves the performance of the overall system.

[0018] Preferably, the mission information includes a single preset-time for all movement phases or an individual preset-time for each movement phase. With this approach the preset-time can be adapted more flexibly to the tasks or environmental situation, e.g., traffic distance between waypoints, etc.

[0019] Preferably, the number of mobile agents includes a leader agent, and for all remaining mobile agents, a proportional delay factor is determined such that the travel times of the leader agent and the remaining mobile agents from the starting location to the destination within the respective phase are identical. With this approach the actions of the mobile agents are adapted to a single mobile agent. A more coordinated performance of the tasks and phases is expected.

[0020] Preferably, the control commands are generated to coordinate the mobile agents by determining, for each mobile agent, a normalized acceleration based on arrival times of neighboring mobile agents and an auxiliary function that depends on the presettime. By controlling the acceleration of each mobile agent, and thereby its current velocity and position, each group of mobile agents can be coordinated to arrive at the same time at its destination.

[0021] Preferably, the normalized acceleration is determined according to where vs^p>it) designates the acceleration between starting location s and destination p for mobile agent number i at time t, Cspdesignates the euclidean distance between starting location s and destination p, <p(t) designates the auxiliary function and <p(t) its derivative with respect to time t, ^(t) designates a correction that depends on neighboring mobile agents, and y and s are tuning parameters. With this equation for the acceleration, the coordination between the mobile agents is improved.

[0022] Preferably, the auxiliary function is chosen as where parameter a > 2 is a free parameter. With this auxiliary function, it can be assured that all requests regarding pickup and drop off locations can be respected within the given time window T.

[0023] Preferably, the correction is chosen as where nonzero atdenotes a weighted gain and :N designates the set of neighboring agents such that > 0. With this correction, only neighboring mobile agents influence each other. Preferably, the neighboring agents are chosen such that their communication graph forms a directed spanning tree.

[0024] The invention provides a mobile robot, a vehicle or a data processing apparatus comprising means for carrying out the previously described method. The invention provides a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the previously described method.

[0025] The invention provides a computer-readable medium or a data carrier signal comprising the computer program.

[0026] This technology disclosure aims to provide a solution for coordinating multiple mobile agents using a preset-time based coordination method. Preset-time based coordination means that a group of interacting agents (robots or vehicles) can cooperatively complete specific tasks within a preset-time window.

[0027] In contrast to traditional centralized or consensus-based methods, preset-time coordination methods hold advantages such as flexible task management, strong system performance guarantee and higher autonomy. This approach can have a wide range of applications in wireless sensor networks, multi-robot systems and smart transportation systems, etc.

[0028] Demand for the ideas described herein is driven by rapid development of demanddependent requests in fields such as assembly line logistics and intelligent logistics where package deliveries using robots are of great importance. Further applications include the transportation of individuals or a group of persons.

[0029] Typically, passengers or packages should be picked up at multiple different depots, be delivered to different destinations, and finally the mobile agents, e.g., vehicles or robots, should return to depots.

[0030] In these scenarios, the missions usually have a respective time window constraint for mission completion. This disclosure mainly concentrates on a delivery problem where all passengers or packages are cooperatively picked up and delivered within a preset-time window. It proposes a preset-time coordination method to achieve package delivery, by which missions with different requirements (pickup / delivery at a common or different locations) can be completed in a unified manner. Preset-time coordination typically involves agents cooperatively completing tasks within a preset-time window. Specifically, preset-time coordination may involve a group of n interacting mobile agents that can be denoted by a set N = {1 , n} and a specified time window T (also designated as „preset-time T“ for short), whose purpose is to cooperatively complete specified tasks within the specified time window T. The task may involve transporting goods or individuals from a pick-up location to a destination.

[0031] In some embodiments, a sequence of transport requests, e.g., package delivery requests or person transport requests, need to be met, in the sense that the respective transported packages or individuals are required to be cooperatively picked up at depots and delivered to specific destinations, within specified time windows, respectively, for pickup and delivery phases.

[0032] In some embodiments, a group of interacting mobile agents (e.g., mobile robots or vehicles), may be used to cooperatively complete pickup and delivery / transport missions within a specified time window. Each of the mobile agents can be assigned to a pickup / delivery task that may be associated with the same (origin) depot.

[0033] The coordination method is developed such that preferably the mobile agents (mobile robots or vehicles) are capable for pickup and delivery tasks and the time window constraints of pickup and delivery phases are respected. For possible pickup and delivery / transport tasks, multiple phases can be defined.

[0034] In some embodiments an active phase, a handshaking phase and a bye-bye phase can be used. The phases can be determined by three specific requirements: a coordination task (picked / delivered to the same location or different locations), a time window constraint (tasks should be completed within a preset time), and mobile agents (robots or vehicles) return to depots. The coordination task characterizes the active phase, the time window constraint characterizes the handshake phase and the mobile agents' return characterizes the bye-bye phase.

[0035] The method can be described by

[0036] 1 ) Input: Locations of depots, delivery points and time window for each phase 2) Output: Schedule policy

[0037] The active phase may process the number of mobile agents n; the communication topology Gi; the actual coordination method, and a first preset Time Ti. The handshaking phase may process again the number of mobile agents n; the communication topology G2; the coordination method; and a second preset-time T2. The bye-bye phase may process again the number of mobile agents n; the communication topology G3; the coordination method; and a third preset-time T3.

[0038] In some embodiments, in the active phase, upon giving the delivery / transport destinations, the number of mobile agents and the locations of the original depots, a coordination algorithm, which is specified by preset-time constraint and a communication topology, are proposed to describe the agents’ dynamic equation where model and parameter are needed to be adjusted if certain error occurs. By such a manner, robots cooperatively can arrive at the delivery points. For the handshaking phase, upon giving the transportation points, the number of robots and the locations of the final depots, another preset-time coordination algorithm is proposed to govern the agents’ dynamic equation, within certain tolerable error. By such a manner, robots cooperatively can return to depots eventually. In summary, the active phase guarantees that starting from the original depots, robots can cooperatively arrive at delivery points; while the handshaking phase guarantees that starting from the delivery points, robots can cooperatively arrive at the final depots, thus completing the delivery / transport mission. Finally, the bye-bye phase may guarantee that the mobile agents can cooperatively return to their respective depots.

[0039] BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Embodiments of the invention are described in more detail with reference to the accompanying schematic drawings.

[0041] Fig. 1 depicts an embodiment of a method according to the invention; and Fig. 2 depicts a schematic result of a verification simulation. It should be noted that the invention is described with reference to a package delivery scenario, but the ideas described herein may well be applied to passenger transport in vehicles.

[0042] DETAILED DESCRIPTION OF EMBODIMENT

[0043] According to the Fig., a method for operating multiple mobile agents 26, such as mobile robots or vehicles, includes an active phase 10, a handshake phase 12 and a bye-bye phase (not shown). The bye-bye phase is not shown for sake of brevity, since it is similar to both the active phase 10 and the handshake phase 12 as described later.

[0044] In the active phase 10, mission information 14 is generated. The mission information 14 includes a delivery destination and a number of mobile robots as mobile agents 26. Furthermore, the mission information 14 may include the starting locations 16 or location of original depot of each mobile agent 26.

[0045] The active phase 10 may further include a preparation step 18. The preparation step 18 may comprise selection of a preset-time constraint Ti. The preset-time constraint Ti can be determined based on the delivery destination. The preset-time constraint may correspond to a time window in which the mobile agents 26 must cooperatively complete the mission. Furthermore, the preparation step determines a communication topology L . The communication topology is indicative of which mobile agents 26 can communicate with each other. In particular, the communication topology includes information about the nearest neighboring mobile agents 26 of each mobile agent 26.

[0046] The preset-time constraint Ti and the communication topology are fed into a delivery-space dynamic model 20. The delivery-space dynamic model 20 is indicative of the movement of each mobile agent 26 between its respective starting location 16 (i.e. , its associated depot) and its destination (i.e. , the pickup location).

[0047] Initially the delivery-space dynamic model 20 determines a Euclidean distance Cspbetween the starting location 16 and the destination for each mobile agent 26. The distance can be retrieved from map data of the environment, for example. The delivery-space dynamic model 20 may arbitrarily select one of the mobile agents 26 as a leader agent. Preferably, the leader agent is the one having the largest value of Euclidean distance Csp. The leader agent is typically designated i = 1 and gets assigned a proportional delay factor of = 1. The remaining mobile agents 26 get associated a proportional delay factor 5,- that can be estimated using the respective euclidean distance and typical velocity of the respective mobile agent 26.

[0048] The delivery-space dynamic model 20 includes the following equations: where vs^(t) designates the acceleration between starting location s and destination p for mobile agent 26 number i at time t, Cspdesignates the euclidean distance between starting location s and destination p, <p(t) designates the auxiliary function and <p(t) its derivative with respect to time t, ^(t) designates a correction that depends on neighboring mobile agents 26, and y and s are tuning parameters.

[0049] The auxiliary function is chosen as where parameter a > 2 is a free parameter and T is the preset-time Ti.

[0050] The correction is chosen as where nonzero atdenotes a weighted gain and :N designates the set of neighboring agents such that

[0051] It should be noted that in equations (1) and (2) the velocity is typically a vector and may include components indicative of the position of the respective mobile agent 26 as well as its orientation.

[0052] For each mobile agent 26 the equation (1) is solved using numerical integration techniques. Thereby the time evolution of the agent position, agent velocity and / or agent acceleration are obtained. These data are transformed into control commands 24 for each mobile agent 26 to cause movement of the mobile agent 26 in accordance with mission information 14 from the starting location 16 to the destination. The control command 24 is executed by the mobile agent 26.

[0053] The active phase 10 includes a check step 28 which checks whether the mobile agent 26 has reached its destination within an acceptable tolerance. If this is not the case, the parameters of the delivery-space dynamic model 20 are optimized and equation (1) is solved again. If the mobile agent 26 has reached its destination, then the package is picked up and the handshake phase 12 is started.

[0054] The handshake phase 12 is similar to the active phase 10 and will therefore only be described in its differences. In the handshake phase 12, the pickup location is now the starting location 16 and the destination is the drop off location. Similarly, the handshake phase 12 includes a check step 28 which checks whether the mobile agent 26 has reached its destination within an acceptable tolerance. If this is not the case, the parameters of the delivery-space dynamic model 20 are optimized and equation (1) is solved again. If the mobile agent 26 has reached its destination, then the package is dropped off and the bye-bye phase 30 is started.

[0055] The bye-bye phase 30 is similar to the active phase 10 and the handshake phase 12 and will therefore only be described in its differences. In the bye-bye phase 30, the drop off location is now the starting location 16 and the destination is the depot, in which the mobile agent originally was housed. Again, a check step checks whether the mobile agent 26 has reached its depot within an acceptable tolerance. If this is not the case, the parameters of the delivery-space dynamic model 20 are optimized and equation (1) is solved again. If the mobile agent 26 has reached its destination, the system waits for the next mission request which initiation mission information 14 generation.

[0056] To verify the effectiveness of the proposed preset-time coordination method for the dial-a-ride problem (DARP) of interest, the inventors considered a DARP described by: a fleet of twelve mobile agents (k = 12) are originally located in several depots; they are required to arrive at 4 points (pickup locations), and then arrive at 2 different locations (destinations); and all mobile agents return to a common depot eventually. It is worthwhile to illustrate that for the DARP just formulated, it is not explicitly specified how many requests are demanded to be collected. In fact, as all requests are known in prior, it is naturally assumed that k mobile agents are sufficient to meet the transportation requests, and in the meanwhile, respecting the capacity of each mobile agent.

[0057] In the following, the communication graph for participating mobile agents will be defined with a = 1 if there is an edge from node j to node i; and 0 otherwise. The used proportional delay factors (PDFs) involving the three phases are chosen as

[0058] 3- = {1, 1, 1, 2, 2, 2, 3,3, 3, 4, 4, 4}

[0059] 3- = {1.5, 1.5, 1.5, 2, 2, 2, 2, 2, 2, 1.5, 1.5, 1.5}

[0060] 8 = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}

[0061] Weighted gains are chosen by aj = {-0.1, 1.1, 0.95, 1.25, 1, 0.8, 1, 1.35, 0.75, 1, 1.5, 1.1} a? = {-o.l, 1.1, 0.95, 1.25, 1, 0.8, 1, 1.35, 0.75, 1, 1.5, 1.1} a? = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}

[0062] No common signs among PDFs and weighted gains are employed. Moreover, the preset-times for the considered three phases are

[0063] T1= 3; T2= 2.5; T3= 2 and a1= 3; a2= 3; a3= 3 e1= 1 ; e2= 1 ; e3= 1 Y1= 1.25; Y2= 1.25; Y3= 1.25

[0064] The preset-times T1 , T2 and T3 are cooperative preset-times which all agents must cooperatively meet. To perform the verification, the initial conditions of the agents are randomly selected from interval [-10; 10], By leverage of the preparations just discussed, the trajectories of the agents are depicted in Fig. 2.

[0065] From Fig. 2, we can observe that the time horizon in each phase is specified by 5. In active phase 10, mobile agents can cooperatively arrive at 4 pick-up locations at time 3, and the corresponded duration is 2; while in handshaking phase 12, mobile agents can cooperatively arrive at 2 drop-off locations at time 5+2.5, and the corresponded duration is 2:5; analogously, in Bye-bye phase 30, all agents cooperatively return a common depot at time 10+2, thus achieving the specified DARP.

[0066] From Fig. 2, we can see that the agents may be re-grouped after having completed a group task within preset-time T 1 and may start another group task (with preset-time T2) as a member of another group. Furthermore, all the agents may also be grouped together in a single group (with preset-time T3).

[0067] With the help of Fig. 2, we can draw a conclusion that the proposed preset-time coordination algorithm in this paper is capable for DARPs at which cooperative pickup and drop-off tasks are needed. A byproduct uncovered by Fig. 5 is: In Handshaking phase 12, final aggregated trajectories of the vehicles do not entirely preserve by the convex hull spanned by the initial conditions associated with the corresponding phase.

[0068] REFERENCE SIGNS

[0069] 10 active phase

[0070] 12 handshake phase

[0071] 14 mission information

[0072] 16 starting location

[0073] 18 preparation step

[0074] 20 delivery-space dynamic model

[0075] 24 control command

[0076] 26 mobile agent

[0077] 28 check step

[0078] 30 bye-bye phase

Claims

CLAIMS1 . A computer-implemented method for operating a plurality of mobile agents (26) in an environment, the environment comprising a set of depots, a set of pickup locations and a set of drop off locations, wherein the method comprises: a) generating mission information (14) that is indicative of any of a number of mobile agents (26), a pickup location, a drop off location, and at least one preset-time, wherein each mobile agent (26) is associated with a depot, wherein the preset-time is indicative of the maximum time period in which the number of mobile agents (26) has to cooperatively complete a movement phase; b) at least one movement phase including, for each of the number of mobile agents (26), generating a control command (24) based on the mission information (14), wherein the control command (24) causes the respective mobile agent (26) to move from a starting location (16) to a destination in accordance with the mission information (14); wherein each control command (24) is generated such that the respective mobile agent (26) arrives at its destination within the preset-time.

2. The method according to claim 1 , wherein in step b) the control command (24) is generated such that, after the collective arrival of the mobile agents (26) at the destination, the mobile agents (26) perform a task.

3. The method according to any of the preceding claims, wherein in step b) a first movement phase called an active phase (10) includes, for each of the number of mobile agents (26), generating a control command (24) based on the mission information (14), wherein the control command (24) causes the respective mobile agent (26) to move from its depot as the starting location (16) to a pickup location as its destination in accordance with the mission information (14), wherein each control command (24) is generated such that the respective mobile agent (26) arrives at the pickup location within the preset-time.

4. The method according to any of the preceding claims, wherein in step b) a second movement phase called a handshaking phase (12) includes, for each of the number of mobile agents (26), generating a control command (24) based on the mission information (14), wherein the control command (24) causes the respective mobileagent (26) to move from a pickup location to a drop off location in accordance with the mission information (14), wherein each control command (24) is generated such that the respective mobile agent (26) arrives at the pickup location within the presettime.

5. The method according to any of the preceding claims, wherein in step b) a third movement phase called a bye-bye phase (30) includes, for each of the number of mobile agents (26), generating a control command (24) based on the mission information (14), wherein the control command (24) causes the respective mobile agent (26) to move from a drop off location back to its depot, wherein each control command (24) is generated such that the respective mobile agent (26) arrives at its depot within the preset-time.

6. The method according to any of the preceding claims, wherein the control command (24) is generated such that, after the collective arrival of the mobile agents (26) at the pickup location, the mobile agents (26) collectively pick up a package or a passenger; and / or wherein the control command (24) is generated such that, after the collective arrival of the mobile agents (26) at the pickup location, the mobile agents (26) collectively transport a package or a passenger to the drop off location; and / or wherein the control command (24) is generated such that, after the collective arrival of the mobile agents (26) at the drop off location, the mobile agents (26) collectively drop off a package or a passenger and the mobile agents (26) return to their respective depots.

7. The method according to any of the preceding claims, wherein the mission information (14) includes a single preset-time for all movement phases or an individual preset-time for each movement phase.

8. The method according to any of the preceding claims, wherein the number of mobile agents (26) includes a leader agent, and for all remaining mobile agents (26), a proportional delay factor is determined such that the travel times of the leader agent and the remaining mobile agents (26) from the starting location (16) to the destination within the respective phase are identical.

9. The method according to any of the preceding claims, wherein the control commands (24) are generated to coordinate the mobile agents (26) by determining, for each mobile agent (26), a normalized acceleration based on arrival times of neighboring mobile agents (26) and an auxiliary function that depends on the presettime.

10. The method according to claim 9, wherein the normalized acceleration is determined according towhere vs^p>it) designates the acceleration between starting location s and destination p for mobile agent number i at time t, Cspdesignates the euclidean distance between starting location s (16) and destination p, <p(t) designates the auxiliary function and <p(t) its derivative with respect to time t, 2,( designates a correction that depends on neighboring mobile agents (26), and y and s are tuning parameters.11 . The method according to claim 9 or 10, wherein the auxiliary function is chosenwhere parameter a > 2 is a free parameter.

12. The method according to claim 10 or 11 , wherein the correction is chosen aswhere nonzero atdenotes a weighted gaindesignates the set of neighboring agents such that13. A mobile robot, a vehicle or a data processing apparatus comprising means for carrying out the method according to any of the preceding claims.

14. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method according to any of the claims 1 to 12.

15. A computer-readable medium or a data carrier signal comprising the computer program according to claim 14.

Images