Predictive path coordination in multi-robot systems
A system and methods for operating a multi-robot system (MRS) are disclosed. An example method can include receiving at least one transportation task; determining an optimal path for executing the at least one transportation task based at least in part on: (i) one or more transportation task parameters, (ii) a shared global critic function accessible to the first robot and the at least one additional robot, and (iii) a local critic function unique to the first robot; and executing the at least one transportation task in accordance with the determined optimal path.
1 . A multi-robot system comprising:
an edge server; and
a plurality of robots in electronic communication with one another and the edge server, each robot being configured to:
receive at least one transportation task from the edge server;
determine an optimal path for executing the at least one transportation task based at least in part on: (i) one or more transportation task parameters, (ii) a shared global critic function accessible to the plurality of robots, and (iii) a local critic function unique to the respective robot, and
execute the at least one transportation task in accordance with the determined optimal path, wherein each of the plurality of robots is configured to initiate a token-passing based resolution to resolve potential collisions with one another in real-time.
2 . The multi-robot system of claim 1 , wherein each of the plurality of robots is configured to execute a plan switch in response to detecting a deviation from a path plan or deadlock with another robot.
3 . The multi-robot system of claim 1 , wherein the optimal path is determined based at least in part on a multi-floor graph-based model.
4 . The multi-robot system of claim 3 , wherein each of the plurality of robots is configured to determine the optimal path by applying a random walk-based path embedding algorithm to the multi-floor graph-based model.
5 . The multi-robot system of claim 1 , wherein the one or more transportation task parameters include at least one of a pickup vertex, a delivery vertex, a target pickup time, a target delivery time, and energy efficiency.
6 . The multi-robot system of claim 1 , wherein each of the plurality of robots is configured to:
continuously determine global states and global rewards based at least in part on the shared global critic function; and
continuously determine local states and local rewards based at least in part on its respective local critic function, wherein the determined global states, global rewards, local states, and local rewards are used to determine or modify the optimal path.
7 . The multi-robot system of claim 1 , wherein the shared global critic function is configured to facilitate predicting a global reward corresponding to a contribution of a pickup or delivery path.
8 . The multi-robot system of claim 1 , wherein each of the plurality of robots comprises one or more unsupervised machine learning models or deep learning models.
9 . The multi-robot system of claim 8 , wherein each of the plurality of robots is configured to host and continuously train the one or more unsupervised machine learning models or deep learning models.
10 . The multi-robot system of claim 8 , wherein the one or more unsupervised machine learning models or deep learning models are trained using randomly generated and embedded paths.
11 . The multi-robot system of claim 1 , wherein the shared global critic function comprises a neural network model.
12 . The multi-robot system of claim 1 , wherein each of the plurality of robots is configured to:
receive a global reward in response to executing a pickup or delivery within a target time frame and receive a diminished global reward in response to failing to execute the pickup or delivery within the target time frame.
13 . A method of operating a first robot in a multi-robot system, wherein the first robot and at least one additional robot of the multi-robot system are in electronic communication with one another, the method comprising:
receiving, by the first robot, at least one transportation task;
determining, by the first robot, an optimal path for executing the at least one transportation task based at least in part on: (i) one or more transportation task parameters, (ii) a shared global critic function accessible to the first robot and the at least one additional robot, and (iii) a local critic function unique to the first robot; and
executing, by the first robot, the at least one transportation task in accordance with the determined optimal path.
14 . The method of claim 13 , further comprising:
initiating, by the first robot, a token-passing based resolution to resolve potential collisions with the at least one additional robot in real-time.
15 . The method of claim 13 , further comprising:
in response to detecting a deviation from a path plan or deadlock with another robot, executing, by the first robot, a plan switch.
16 . The method of claim 13 , wherein the first robot is configured to determine the optimal path by applying a random walk-based path embedding algorithm to a multi-floor graph-based model.
17 . The method of claim 13 , wherein the first robot is further configured to:
train one or more unsupervised machine learning models or deep learning models using randomly generated and embedded paths.
18 . The method of claim 13 , wherein the one or more transportation task parameters include at least one of a pickup vertex, a delivery vertex, a target pickup time, a target delivery time, and energy efficiency.
19 . The method of claim 13 , wherein the first robot is further configured to:
continuously determine global states and global rewards based at least in part on the shared global critic function; and
continuously determine local states and local rewards based at least in part on the local critic function, wherein the determined global states, global rewards, local states, and local rewards are used to determine or modify the optimal path.
20 . A non-transitory computer-readable medium comprising a memory having instructions stored thereon to cause a processor to:
receive at least one transportation task;
determine an optimal path for executing the at least one transportation task via at least one robot based at least in part, on: (i) one or more transportation task parameters, (ii) a shared global critic function accessible to a plurality of robots, and (iii) a local critic function unique to each of the plurality of robots, and
cause execution of the transportation task in accordance with the determined optimal path, wherein at least one robot is configured to initiate a token-passing based resolution to resolve potential collisions with at least another robot in real-time.