Page 66 - Read Online

Page 6 of 15 Ma et al. Intell Robot 2023;3(4):581-95 I http://dx.doi.org/10.20517/ir.2023.33

( , )=1, it indicates that agent is assigned to execute task , otherwise ( , )=0.

The constraints of the Assignment problem include the following aspects:

Task constraint: Each task can only be executed by one agent.

∑
( , ) = 1,for all tasks = 1, 2, (2)
=1

Agent constraint: Each agent can execute at most one task.

∑
( , ) ≤ 1,for all agents = 1, 2, (3)
=1

Resource constraints: Each intelligent agent needs to comply with its own resource constraints when per-
forming tasks, such as considering battery energy for drones and unmanned vehicles and fuel constraints for
unmanned ships.

Task type constraint: Some tasks may require specific types of intelligent agents to execute; for example, only
drones can perform aerial reconnaissance tasks, and only unmanned vehicles can perform tasks on land.

Avoiding collision constraints: In order to ensure that there is no collision between intelligent agents, it is
necessary to plan and coordinate the paths of intelligent agents to avoid their intersection.

In practical applications, this assignment problem may require real-time and adaptive performance. Because
the environment is unknown and tasks may change at any time, intelligent agents need to be able to make
decisions and adjustments based on real-time information.

3. HETEROGENEOUS AGENT TASK ALLOCATION ALGORITHM BASED ON GRAPH HETERO-

GENEOUS NEURAL NETWORK ANT COLONY OPTIMIZATION ALGORITHMS
The GHNN-ACO algorithms proposed in this paper are a hybrid algorithm integrating GHNN and ACO
algorithms. These algorithms are mainly used to address the problems of node classification, link prediction,
and graph feature learning in the heterogeneous data of a multi-agent graph.

3.1. Improvement and application of GHNN-ACO in heterogeneous multi-agent emergency rescue
mission scenarios
In the context of a heterogeneous multi-agent economic rescue mission, we will refine the application of ACO
algorithms to a heterogeneous neural network. Let us assume that a natural disaster requires emergency eco-
nomicrescue. Wehavemultipleintelligentagents, includingdrones, unmannedships, andunmannedvehicles,
with various tasks such as searching for and rescuing disaster victims, transferring affected residents, and dis-
tributing rescue supplies. Our objective is to maximize overall efficiency by intelligently allocating tasks and
facilitating collaboration within limited resources.

In a specific multi-agent task scenario, the first step is initialization: setting the number of agents, the number
of tasks, the structure of the graph for the heterogeneous neural network, the parameters of ACO algorithms,