Coalition Formation for Multiple Heterogeneous UAVs in Unknown Environment

A Survey of UAV Swarm Task Allocation Based on the Perspective of Coalition Formation

“…In addition to the auction algorithm, the contract network algorithm, and their improvement methods, other distributed algorithms have been applied to solve the formation of UAV swarm coalitions. Zheng et al (2022) proposed a distributed coalition formation method based on a Monte Carlo tree search for the distributed coalition formation problem of heterogeneous UAV swarms in unknown dynamic environments. This method designed a coalition task automaton and optimized the coalition structure through a two-stage Monte Carlo tree search, which can effectively seek a solution for large-scale distributed coalition formation problems.…”

A Coalition Formation Approach to Coordinated Task Allocation in Heterogeneous UAV Networks

“…It should be noted that the k'th leader is a member of the coalition S k . In order to find the optimal coalition which maximizes the leader's utility function (10), a search over all 2 L k possible coalitions is required, where L k denotes the number of potential follower UAVs who responded to the proposal of leader k. To avoid such extensive search, a low complexity mergeand-split algorithm is proposed. In this method, each leader k separately starts from an initial state where the set of UAVs who responded to its proposal is partitioned into L k singleton coalitions.…”

“…Also, if for a nonsingleton coalition S there exists a partition of two coalitions S ki and S kj such that v(S = S ki S kj ) < v(S ki ) + v(S kj ), then S splits into S ki and S kj 3 . At each step of the merge and split algorithm, the optimum value of the SN R B in the utility function (10) for the coalitions should be calculated, as described in details in the next section. It is worth mentioning that the merge and split algorithm is used to obtain a suboptimal coalition solution of (10) with lower complexity.…”

“…The optimization from the leaders' perspective is to maximize the coalition value (10) by searching over the coalitions, i. e. S, and determine the optimum beamforming scheme, i. e. w to optimize SN R B . For each coalition, the optimal value of (5), SN R opt S can be obtained via the following inner optimization problem:…”

“…The task allocation problem in multi-agent systems is defined as the process of allocating a set of tasks to groups of agents to ensure timely and efficient task completions, noting the individual capabilities of the agents [3]. Several studies have addressed the complex problem of task assignment using different approaches including mixed integer linear programming [4], [5], dynamic network flow optimization [6], [7], market-based strategy [8], [9], finite state machine [10], and multiple choice knapsack problem [11]. While in the majority of the aforementioned works, the tasks are centrally assigned to the agents by a base station that has complete knowledge about the tasks and often the agents' capabilities; in many dynamic systems, tasks may appear at unpredictable locations and times (e.g.…”