Page 65 - Read Online

Ma et al. Intell Robot 2023;3(4):581-95 I http://dx.doi.org/10.20517/ir.2023.33 Page 5 of 15

Table 1. A binary variable of task allocation
A binary variable X (i, j) Definition
( , ) = 1 Agent is assigned to execute task
( , ) = 0 Agent does not execute task

to swiftly obtain images and information of the affected area. Personnel search and rescue tasks are assigned
to unmanned ships and unmanned vehicles, which use underwater sensors for search and rescue missions in
water, while unmanned vehicles conduct personnel search and rescue on the ground. Unmanned vehicles are
responsible for handling the material transportation task, efficiently transporting rescue materials from the
command center to the disaster area using their off-road capabilities and load-carrying ability.

Path planning and conflict avoidance: UAVs, unmanned ships, and unmanned vehicles utilize Fast path plan-
ning algorithms, such as A* algorithm or RRT algorithm, to plan the optimal path based on task urgency and
agent speed. During task execution, sensors are used to perceive the surrounding environment in real time,
and obstacle avoidance algorithms are employed to prevent collisions with other intelligent agents or obstacles.

Task execution monitoring and adjustment: The intelligent agents periodically report the task execution sta-
tus and progress to the emergency rescue command center, including the percentage of completed tasks and
encounteredproblems. Theemergencyrescuecommandcenteradjuststaskallocationandcollaborationstrate-
gies in real time based on the reports from the intelligent agents and the arrival of new tasks. The most urgent
tasks are given priority, and the most suitable intelligent agents are scheduled to maximize rescue efficiency.

2.2. Heterogeneous multi-agent task scenarios
In the heterogeneous multi-agent task allocation scenario of this article, we have three types of agents: UAVs,
unmannedships(USVs), andunmannedvehicles(AVs). Theyaredeployedinanunknownenvironment, such
as a remote exploration area or an unknown sea area. The task objective is to perform a series of collaborative
tasks in an unknown environment, such as exploration, target search, data collection, etc., to achieve specific
task objectives. Each intelligent agent has its own resource limitations, such as battery energy (for drones and
unmanned vehicles), fuel(forunmanned ships), andcomputingpower. Taskscanbe divided into investigating
targets, collecting data, transporting items, and so on.

We model the assignment problem in the scenario as a multi-agent optimization problem. Assuming we have
n intelligent agents, n represents the number of drones, unmanned ships, and unmanned vehicles, respectively.
We use to represent the -th agent and to represent the -th task. Each agent can execute task , and each
task has a specific reward or value ( , ).

To represent task allocation, we introduce a binary variable ( , ), which is defined as follows:

( , ) = 1, if agent is assigned to execute task ;

( , ) = 0, if agent does not execute task .

Our goal is to maximize the effectiveness of the entire team, which is the sum of the total rewards for tasks
performed by all intelligent agents. It can be expressed as:

∑ ∑
Maximize ( , ) × ( , ) (1)
=1 =1

Among them, ( , ) is a binary variable that represents whether agent is assigned to execute task . When