Market or auction based algorithms offer effective methods for de-centralized task assignment in multi-agent teams. Typically there is an implicit assumption that agents are willing to cooperate and can be trusted to perform assigned tasks. Reciprocal collaboration may not always be a valid assumption. In cases where auctions are used for task allocation, without explicit revenue exchange, incentives are needed to enforce cooperation. An approach to incentive based trust is presented, which enables detection of team members that are not contributing and for dynamic formation of teams.

Game theoretic predictions about equilibrium behavior depend upon assumptions of inflexibility of belief, of accord between belief and choice, and of choice across situations that share a game-theoretic structure. However, researchers rarely possess any knowledge of the actual beliefs of subjects, and rarely compare how a subject behaves in settings that share game-theoretic structure but that differ in other respects. Our within-subject experiments utilize a belief elicitation mechanism, roughly similar to a prediction market, in a laboratory setting to identify subjects’ beliefs about other subjects’ choices and beliefs. These experiments additionally allow us to compare choices in different settings that have similar game-theoretic structure. We find first, as have others,that subjects’ choices in the Trust and related games are significantly different from the strategies that derive from subgame perfect Nash equilibrium principles. We show that, for individual subjects, there is considerable flexibility of choice and belief across similar tasks and that the relationship between belief and choice is similarly flexible. To improve our ability to predict human behavior, we must take account of the flexible nature of human belief and choice

There are an increasing number of applications of security games. One of the key challenges for this field going forward is to address the problem of model uncertainty and the robustness of the game-theoretic solutions. Most existing methods for dealing with payoff uncertainty are Bayesian methods which are NP-hard and have difficulty scaling to very large problems. In this work we consider an alternative approach based on interval uncertainty. For a variant of security games with interval uncertainty we introduce a polynomial-time approximation algorithm that can compute very accurate solutions within a given error bound.

In some urban transit systems, passengers are legally required to purchase tickets before entering but are not physically forced to do so. Instead, patrol units move about through the transit system, inspecting tickets of passengers, who face fines for fare evasion. This setting yields the problem of computing optimal patrol strategies satisfying certain temporal and spacial constraints, to deter fare evasion and hence maximize revenue. In this paper we propose an initial model of this problem as a leader-follower Stackelberg game. We then formulate an LP relaxation of this problem and present initial experimental results using real-world ridership data from the Los Angeles Metro Rail system.

This paper considers the problem of patrolling multiple targets in a Euclidean environment by a single patrolling unit. We use game-theoretic approach and model the problem as a two-player zero-sum game in the extensive form. Based on the existing work in the domain of patrolling we propose a novel mathematical non-linear program for finding strategies in a discretized problem, in which we introduce a general concept of internal states of the patroller. We experimentally evaluate game value for the patroller for various graphs and strategy representations. The results suggest that adding internal states for the patroller yields better results in comparison to adding choice nodes in the used discretization.

Successfully accomplishing everyday manipulation tasks requires robots to have substantial knowledge about the objects they interact with, the environment they operate in as well as about the properties and effects of the actions they perform. Often, this knowledge is implicitly contained in manually written control programs, which makes it hard for the robot to adapt to newly acquired information or to re-use knowledge in a different context. By explicitly representing this knowledge, control decisions can be formulated as inference tasks which can be sent as queries to a knowledge base. This allows the robot to take all information it has at query time into account to generate answers, leading to better flexibility, adaptability to changing situations, robustness, and the ability to re-use knowledge once acquired. In this paper, we report on our work towards a practical and grounded knowledge representation and inference system. The system is specifically designed to meet the challenges created by using knowledge processing techniques on autonomous robots, including specialized inference methods, grounding of symbolic knowledge in the robot's control structures, and the acquisition of the different kinds of knowledge a robot needs.

Evolutionary language games are a way to study how perceptions, concepts, and language can emerge in populations of situated embodied agents, driven by the needs of communication and the properties of the environment. Evolutionary language games are currently being investigated using physical robots and this then requires that the full cycle of processing activities from physical robotic embodiment to sensory-motor processing, visual perception and action, conceptualization, and language processing are all integrated in a single system. This contribution reports on a large-scale long term effort to experiment with evolutionary language games and discusses major results achieved so far.

Constructing robust controllers to perform tasks in large, continually changing worlds is a difficult problem. A long-lived agent placed in such a world could be required to perform a variety of different tasks. For this to be possible, the agent needs to be able to abstract its experiences in a reusable way. This paper addresses the problem of online multitask decision making in such complex worlds, with inherent incompleteness in models of change. A fully general version of this problem is intractable but many interesting domains are rendered manageable by the fact that all instances of tasks may be described using a finite set of qualitatively meaningful contexts. We suggest an approach to solving the multitask problem through decomposing the domain into a set of capabilities based on these local contexts. Capabilities resemble the options of hierarchical reinforcement learning, but provide robust behaviours capable of achieving some subgoal with the associated guarantee of achieving at least a particular aspiration level of performance. This enables using these policies within a planning framework, and they become a level of abstraction which factorises an otherwise large domain into task-independent sub-problems, with well-defined interfaces between the perception, control and planning problems. This is demonstrated in a stochastic navigation example, where an agent reaches different goals in different world instances without relearning.

Natural world interaction (NWI), the pursuit of arbitrary goals in unstructured physical environments, is an excellent motivating problem for the reintegration of artificial intelligence. It is the problem set that humans struggle to solve. At a minimum it entails perception, learning, planning, and control, and can also involve language and social behavior. An agent's fitness in NWI is achieved by being able to perform a wide variety of tasks, rather than being able to excel at one. In an attempt to address NWI, a brain-emulating cognition and control architecture (BECCA) was developed. It uses a combination of feature creation and model-based reinforcement learning to capture structure in the environment in order to maximize reward. BECCA avoids making common assumptions about its world, such as stationarity, determinism, and the Markov assumption. BECCA has been demonstrated performing a set of tasks which is non-trivially broad, including a vision-based robotics task. Current development activity is focused on applying BECCA to the problem of general Search and Retrieve, a representative natural world interaction task.

The fragmentation of research in AI and robotics has created a vast repertoire of skills a robot could be equipped with but that must be manually integrated to form a complex action. We propose a novel evolutionary algorithm that aims at autonomously integrating, adapting and creating new actions by re-using skills that are either externally provided or previously generated. Complex actions are created by instantiating a Finite State Automaton and new skills are created using fully recurrent neural networks. We validated our approach in two scenarios, i.e. exploration and moving to pre-grasp positions. Our experiments show that complex actions can be created by composing independently developed skills. The results have been applied and tested with a real robot in a variety of scenarios.