Sei sulla pagina 1di 1

Appendix

Paper #118
1 APPENDIX
Lemma 4.1. Given a worker w j and her mapping task t π (j) ∈ T −S,
only if the net revenue r π (j) − c π (j), j is no more than zero, there may
be a feasible re-pricing scheme.
Proof. If t π (j) belongs to T − S, it means that the task with the
maximum net revenue, t π (j) , is not selected by the worker w j . So
t π (j) should be a unprofitable task (r π (j) − c π (j), j ≤ 0), otherwise
the linear programming of Eq.(16)-(23) has no solution and the
feasible re-pricing scheme does not exist, because Eq.(21) cannot
be satisfied. 
Lemma 4.2. If the linear programming of Eq.(16)-(23) has solutions
and the equality of Eq.(23) holds, there exists a feasible re-pricing
scheme.
Proof. If the LP has a feasible solution but the equality of Eq.(23)
does not hold, the solution is contrary to the problem setting as
Eq.(4)(5). So only if the LP’s solution can satisfy the equality of
Eq.(23), the solution is a feasible re-pricing scheme. 
Theorem 4.5. The branch-and-bound algorithm is optimal.
Proof. To prove the optimum, the branch-and-bound algorithm
should consider all possible states of the problem and terminate in
finite time.
Completion. A sub-problem is defined to compute a re-pricing
scheme under restricted condition where only a subset of tasks
S are fully completed by all workers W . For the Algorithm 1, it
traverses all kinds of subset S (line 3 and 15), unless the sub-problem
is contrary to the constraint Eq.(6) (line 4 and 16) or hopeless to
exceed the current solution according to Theorem 4.4 (line 11).
For each sub-problem, the Algorithm 2 recursively considers all
possible combinations of worker-task mapping pairs (line 2 and
10-11) and compute whether there is a feasible re-pricing scheme
by the LP of Eq.(16)-(23), unless the branch is pruned due to the
LP’s impossible constraints as Eq.(27)(29). Therefore, the branch-
and-bound algorithm is complete.
Termination. In the Algorithm 1, the number of sub-problems
is finite. And the recursion of Algorithm 2 has an exit to finish, as
line 1. So, the algorithm can terminate. 

Proc. of the 17th International Conference on Autonomous Agents and Multiagent Systems
(AAMAS 2018), M. Dastani, G. Sukthankar, E. Andre, S. Koenig (eds.), July 2018, Stockholm,
Sweden
© 2018 International Foundation for Autonomous Agents and Multiagent Systems
(www.ifaamas.org). All rights reserved.
https://doi.org/doi

Potrebbero piacerti anche