Below you can find some ideas for projects. Students should discuss the ideas presented below with the instructor for the project assignment.

### Constraint Reasoning for Energy Management

Formalise the energy management problem using constraint reasoning. Energy Management aims at reducing peak demand in energy consumption and Constraint Reasoning could be used in many ways. An example is to formalise the energy management problem as a scheduling problem for energy appliances, scheduling on/off periods to reduce peak loads. Another possibility is to aggregate users based on their demand to form coalitions that have an aggregated demand curve that minimises the peak demand.

### Winner determination for Combinatorial Auctions

Implement and evaluate a complete algorithm to solve the winner determination problem for combinatorial auctions. Test the method on randomly generated combinatorial auction instances. Suggested approaches are branch and bound or bucket elimination### Coordination of Sensor Schedules for energy aware networks

Implement and evaluate a complete or approximate algorithm to solve the energy aware wide area surveillance problem. Problem instance can be provided by the instructor. Suggested approaches are Simulated Annealing or other local search methods, Branch and Bound or Bucket Elimination.### Tree Decomposition

Implement the Join Tree Clustering algorithm. Investigate various variable selection orderings and study the ordering impact on the maximal clique size. Test the problem on a MaxCSP version of graph coloring, using benchmarking instances available on the DCOP page (see below).### Coalition formation for Multi Agent Coordination

Implement and evaluate a complete or approximate algorithm to solve the coalition formation problem in Multi-Agent coordination (for a formalisation of this problem as a COP contact the instructor). Test the method on randomly generated instances or on instances provided by the instructor. Suggested solution techniques are: simulated annealing, greedy local search, branch and bound or bucket elimination

- Guidelines for the projects
- DCOPs page Link to the USC DCOP page
- Graph coloring Problem Instances Problem instances for graph coloring as CSP or MaxCSP
- Soft Graph coloring Problem Instances Problem instances for graph coloring and Sensor Network as COP
- ADOPT FAQ ADOPT is a DCOP solver. This page includes a discussion on the format of the graph coloring instances