Partial constraint satisfaction
Artificial Intelligence - Special volume on constraint-based reasoning
On the conversion between non-binary constraint satisfaction problems
AAAI '98/IAAI '98 Proceedings of the fifteenth national/tenth conference on Artificial intelligence/Innovative applications of artificial intelligence
Asynchronous Weak-commitment Search for Solving Distributed Constraint Satisfaction Problems
CP '95 Proceedings of the First International Conference on Principles and Practice of Constraint Programming
Solving Distributed Constraint Optimization Problems Using Cooperative Mediation
AAMAS '04 Proceedings of the Third International Joint Conference on Autonomous Agents and Multiagent Systems - Volume 1
Autonomous Agents and Multi-Agent Systems
Examining DCSP coordination tradeoffs
AAMAS '06 Proceedings of the fifth international joint conference on Autonomous agents and multiagent systems
Asynchronous Forward-checking for DisCSPs
Constraints
Journal of Artificial Intelligence Research
Completeness and performance of the APO algorithm
Journal of Artificial Intelligence Research
Distributed problem solving in geometrically-structured constraint networks
Proceedings of the 2013 international conference on Autonomous agents and multi-agent systems
| Hi-index | 0.00 |
Since its creation, the Asynchronous Partial Overlay (APO) protocol has received a great deal of attention because of its non-traditional approach to solving Distributed Constraint Satisfaction Problems (DCSPs). Its introduction led investigators to question the very definition of the word "distributed" and has subsequently inspired the community to create improved metrics for parallel computation, enhanced testing procedures, and most importantly new DCSP algorithms. These advances have raised concerns about APO's parallel efficiency by showing that, in some cases, APO performs very poorly compared to protocols such as Asynchronous Forward Checking, Conflict-directed Back jumping (AFC-CBJ). In addition, APO's soundness and completeness were brought into question when it was discovered that, under certain conditions, stale state information could cause the protocol's distributed locking mechanism to fail. This work revisits APO by reengineering the protocol to simplify it and increase its parallelism while ensuring its soundness and completeness. It also vastly improves the parallel efficiency of APO by replacing its central solver with a variant of the Forward Checking, Conflict-directed Back jumping (FC-CBJ) algorithm that is specifically tuned to complement the heuristic strategies used by APO to limit its centralization. This new version of APO is then evaluated against the AFC-CBJ protocol using random instances of both DCSPs and distributed 3-coloring problems. The end result is a protocol that is several orders of magnitude faster than the original APO, uses less messages, is more private, and outperforms the AFC-CBJ protocol in nearly every case tested.