CASPIAN JOURNAL
MANAGEMENT AND HIGH TECHNOLOGIES
METHOD FOR TASKS ALLOCATION AND PLANNING THE SEQUENCE OF PERFORMING TASKS BY AGENTS OF SWARM ROBOTIC SYSTEMS UNDER UNCERTAINTY
 |
Read | Petrenko Vyacheslav I., Tebueva Fariza B., Pavlov Andrey S., Gurchinskiy Mikhail M. METHOD FOR TASKS ALLOCATION AND PLANNING THE SEQUENCE OF PERFORMING TASKS BY AGENTS OF SWARM ROBOTIC SYSTEMS UNDER UNCERTAINTY // Caspian journal : management and high technologies. — 2022. — №3. — pp. 25-43. |
Petrenko Vyacheslav I. - North Caucasian Federal University
Tebueva Fariza B. - North Caucasian Federal University
Pavlov Andrey S. - North Caucasian Federal University
Gurchinskiy Mikhail M. - North Caucasian Federal University
The use of swarm robotic systems in a non-deterministic environment actualizes the issues of developing appropriate methods and algorithms for distributing and scheduling tasks. Under the conditions of a non-deterministic environment, we mean such a situation when the maximum number of tasks is limited, and immediately after the first task is completed, a new task appears, that is, the list of tasks changes dynamically during the operation of the swarm robotic systems. When swarm robotic systems operate in a non-deterministic environment, the existing methods and algorithms do not allow optimal distribution of tasks between all robots of the system and plan the sequence of tasks assigned to each of the robots. In addition, the known methods of task distribution and scheduling do not take into account the limitations of the sensory and computing capabilities of robotic devices used in the swarm robotic systems (for example, a small amount of RAM, low processor clock frequency, low battery capacity, low performance of onboard sensors and sensors, etc.). It is also worth noting that not all known methods aimed at solving this problem take into account the specifics of the decentralized control of swarm robotic systems, which consists in a limited scope, because of which their use in real scenarios of using swarm robotic systems is associated with significant problems. The aim of the work is to increase the efficiency of distribution and planning of tasks in the swarm robotic systems in a non-deterministic environment, taking into account the limited capabilities of the elements of the swarm robotic systems and the specifics of decentralized control. The problem was solved using the methods of system analysis, analytical geometry and artificial neural networks. An element of scientific novelty is the proposed algorithms for sorting tasks and searching for transit tasks, which provide an increase in the efficiency of planning and distribution of tasks in swarm robotic systems in a non-deterministic environment, taking into account the limited capabilities of swarm robotic systems elements. The proposed method differs from the known methods by the sorting algorithm of task execution priority in the form of a linked list, which makes it possible to scale the number of swarm robotic systems agents with a dynamic change in the list of urgent tasks. Another difference is the procedure for distributing tasks between swarm robotic systems agents, which makes it possible to search for intermediate tasks to be performed, which reduces the total task execution time compared to similar solutions. Based on the proposed method, the paper presents a neural network modification of this method, which differs by taking into account the specifics of decentralized control. The presented solution is programmatically implemented in Python and can be used in modeling decentralized control systems of swarm robotic systems.
Key words: swarm robotic systems, task allocation, task scheduling, artificial neural networks