Abstract
The use of multiagent systems (MASs) in real-world applications keeps increasing, and diffuses into new domains, thanks to technological advances, increased acceptance, and demanding productivity requirements. Being able to automate the generation of mission plans for MASs is critical for managing complex missions in realistic settings. In addition, finding the right level of abstraction to represent any generic MAS mission is important for being able to provide general solution to the automated planning problem. In this article, we show how a mission for heterogeneous MASs can be cast as an extension of the traveling salesperson problem (TSP), and we propose a mixed-integer linear programming formulation. In order to solve this problem, a genetic mission planner (GMP), with a local plan refinement algorithm, is proposed. In addition, the comparative evaluation of CPLEX and GMP is presented in terms of timing and optimality of the obtained solutions. The algorithms are benchmarked on a proposed set of different problem instances. The results show that, in the presence of timing constraints, GMP outperforms CPLEX in the majority of test instances.
Highlights
M ULTIROBOT systems (MASs) have raised increasing attention in different domains, ranging from underwater [1] to airborne [2] missions
It has been compared to the previous ECTSP formulation (2CFN) by implementing both models in CPLEX, and performing a benchmark on six problem instances, which all represent problems in the domain of mission planning with heterogeneous multiagents with a gradually increasing complexity with respect to problem size
2CFN was more successful in improving suboptimal initial solutions, which were distant from the optimal solution
Summary
M ULTIROBOT systems (MASs) have raised increasing attention in different domains, ranging from underwater [1] to airborne [2] missions. The presented formulations define in a nonambiguous way the problem addressed in this article, and they can be implemented in optimization tools, for example, CPLEX, for computing the optimal solution for the mission planning problem. The main contributions of this article are: 1) a new formulation of the extended CTSP (ECTSP) problem as a MILP problem, with PCs and subtour elimination expressed as an extension of the RMTZ [10] formulation; 2) the proposal of the genetic mission planner (GMP) algorithm, with a local refinement method, for the solution of ECTSP; and 3) a comparative evaluation of the CPLEX implementation of the proposed MILP formulation with a two-commodity flow network (2CFN) formulation [11], [12], and with the GMP approach in terms of solution quality and computational time
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
