Mike Barley

Intelligent Agents and Multi-Agent Systems

We present a reactive agent architecture which incorporates decision-theoretic notions to drive the deliberation and meta-deliberation process, and illustrate how this architecture can be exploited to model an agent who reacts to contextually instantiated norms by monitoring for norm instantiation and replanning its current intentions.

Safety and Security in Multiagent Systems: Research Results from 2004-2006

This paper is concerned with assuring the safety of a swarm of agents (simulated robots). Such behavioral assurance is provided with the physics method called kinetic theory. Kinetic theory formulas are used to predict the macroscopic behavior of a simulated swarm of individually controlled agents. Kinetic theory is also the method for controlling the agents. In particular, the agents behave like particles in a moving gas. The coverage task addressed here involves a dynamic search through a bounded region, while avoiding multiple large obstacles, such as buildings. In the case of limited sensors and communication, maintaining spatial coverage – especially after passing the obstacles – is a challenging problem. Our kinetic theory solution simulates a gas-like swarm motion, which provides excellent coverage. Finally, experimental results are presented that determine how well the macroscopic-level theory, mentioned above, predicts simulated swarm behavior on this task.

Robocup rescue simulation competition: status report

This is the fifth anniversary of the Robocup Rescue Simulation Competitions and the tenth anniversary of the disaster that inspired the Competitions. This is a good time to take stock of what milestones have been achieved and what milestones we should be aiming for. Specifically, this paper looks at the goals that led to the establishment of the competition, the current status of the simulation platform and infrastructure, and finally suggests areas of the current simulation platform which should be improved and parts of the Robocup Rescue technical and social infrastructure which should be extended.

Evaluating the Seeding Genetic Algorithm

In this paper, we present new experimental results supporting the Seeding Genetic Algorithm (SGA). We evaluate the algorithms performance with various parameterisations, making comparisons to the Canonical Genetic Algorithm (CGA), and use these as guidelines as we establish reasonable parameters for the seeding algorithm. We present experimental results confirming aspects of the theoretical basis, such as the exclusion of the deleterious mutation operator from the new algorithm, and report results on GA-difficult problems which demonstrate the SGAs ability to overcome local optima and systematic deception.

Avatars at a meeting

The development of remote avatars has recently generated increased research and commercial interest. Current approaches utilize simple remote-user-guided screens to represent the remote participant. Though humanoid robotic systems are significantly more expensive this work investigates the added benefit from utilizing such a robot. Two recent projects examined the potential of humanoid robotic systems to operate as a remote avatar within a meeting context and their impact on meeting dynamics and interactions. These projects identified the utility of human-like gestures as a significant benefit of humanoid robots within such a setting as well as a range of disruptive impacts due to the operational mode of humanoid robots.

On creating complementary pattern databases

A pattern database (PDB) for a planning task is a heuristic function in the form of a lookup table that contains optimal solution costs of a simplified version of the task. In this paper we introduce a method that sequentially creates multiple PDBs which are later combined into a single heuristic function. At a given iteration, our method uses estimates of the A* running time to create a PDB that complements the strengths of the PDBs created in previous iterations. We evaluate our algorithm using explicit and symbolic PDBs. Our results show that the heuristics produced by our approach are able to outperform existing schemes, and that our method is able to create PDBs that complement the strengths of other existing heuristics such as a symbolic perimeter heuristic.

A New Efficient In Situ Sampling Model for Heuristic Selection in Optimal Search

Techniques exist that enable problem-solvers to automatically generate an almost unlimited number of heuristics for any given problem. Since they are generated for a specific problem, the cost of selecting a heuristic must be included in the cost of solving the problem. This involves a tradeoff between the cost of selecting the heuristic and the benefits of using that specific heuristic over using a default heuristic. The question we investigate in this paper is how many heuristics can we handle when selecting from a large number of heuristics and still have the benefits outweigh the costs. The techniques we present in this paper allow our system to handle several million candidate heuristics.

Safe Stochastic Planning: Planning to Avoid Fatal States

Markov decision processes (MDPs) are applied as a standard model in Artificial Intelligence planning. MDPs are used to construct optimal or near optimal policies or plans. One area that is often missing from discussions of planning under stochastic environment is how MDPs handle safety constraints expressed as probability of reaching threat states. We introduce a method for finding a value optimal policy satisfying the safety constraint, and report on the validity and effectiveness of our method through a set of experiments.

Towards Safe Coordination in Multi-agent Systems

Conservative design is the ability of an individual agent to ensure predictability of its overall performance even if some of its actions and interactions may be inherently less predictable or even completely unpredictable. In this paper, we describe the importance of conservative design in cooperative multi-agent systems and briefly characterize the challenges that need to be addressed to achieve this goal.

Creating better abstract operators

Using abstract operators for least commitment in planning has been shown to potentially reduce the search space by an exponential factor. However a naive application of these operators can result in an unbounded growth in search space for the worst case. In this paper we investigate another important aspect of abstract operators – that of their construction. Similar to their application, naive construction of an abstract operator may leave you with little search space reduction even in the best case, and significant penalties in the worst. We explain what it means to be a good abstract operator and describe a method of creating good abstract operators.