We study the benefit of sharing representations among tasks to enable the effective use of deep neural networks in Multi-Task Reinforcement Learning. We leverage the assumption that learning from different tasks, sharing common properties, is helpful to generalize the knowledge of them resulting in a more effective feature extraction compared to learning a single task. Intuitively, the resulting set of features offers performance benefits when used by Reinforcement Learning algorithms. We prove this by providing theoretical guarantees that highlight the conditions for which is convenient to share representations among tasks, extending the wellknown finite-time bounds of Approximate Value-Iteration to the multi-task setting. In addition, we complement our analysis by proposing multi-task extensions of three Reinforcement Learning algorithms that we empirically evaluate on widely used Reinforcement Learning benchmarks showing significant improvements over the single-task counterparts in terms of sample efficiency and performance. Multi-Task Learning (MTL) ambitiously aims to learn multiple tasks jointly instead of learning them separately, leveraging the assumption that the considered tasks have common properties which can be exploited by Machine Learning (ML) models to generalize the learning of each of them. For instance, the features extracted in the hidden layers of a neural network trained on multiple tasks have the advantage of being a general representation of structures common to each other.

In the Multiagent Connected Path Planning problem (MCPP), a team of agents moving in a graph-represented environment must plan a set of start-goal joint paths which ensures global connectivity at each time step, under some communication model. The decision version of this problem asking for the existence of a plan that can be executed in at most a given number of steps is claimed to be NP-complete in the literature. The NP membership proof, however, is not detailed. In this paper, we show that, in fact, even deciding whether a feasible plan exists is a PSPACE-complete problem. Furthermore, we present three algorithms adopting different search paradigms, and we empirically show that they may efficiently obtain a feasible plan, if any exists, in different settings.

Learning in real-world domains often requires to deal with continuous state and action spaces. Although many solutions have been proposed to apply Reinforcement Learning algorithms to continuous state problems, the same techniques can be hardly extended to continuous action spaces, where, besides the computation of a good approximation of the value function, a fast method for the identification of the highest-valued action is needed. In this paper, we propose a novel actor-critic approach in which the policy of the actor is estimated through sequential Monte Carlo methods. The importance sampling step is performed on the basis of the values learned by the critic, while the resampling step modifies the actor's policy. The proposed approach has been empirically compared to other learning algorithms into several domains; in this paper, we report results obtained in a control problem consisting of steering a boat across a river.

While several studies focused on problems with continuous states, little attention has been deserved to tasks involving continuous actions.

This article reports on the Sixth Robot World Cup Competition and Conference (RoboCup-2002) Fukuoka/Busan, which took place from 19 to 25 June in Fukuoka, Japan. It was the largest Robo- Cup since 1997 and held the first humanoid league competition in the world. Further, the first ROBOTREX (robot trade and exhibitions) was held with about 50 companies, universities, and institutes represented. To the best of our knowledge, this was the largest robotic event in history.

Competitions were held at Since the first competition in 1997 (Kitano Fukuoka Dome Baseball Stadium from 19 to 23 1998), RoboCup has grown into an international June followed by the International RoboCup joint research project in which about Symposium on 24 to 25 June. It is one of RoboCup is an attempt to foster intelligent the most ambitious projects of the twenty-first robotics research by providing a standard century. RoboCup currently consists of three problem, the ultimate goal of which is to divisions: (1) RoboCupSoccer, a move toward build a team of 11 humanoid robots that the final goal; (2) RoboCupRescue, a serious social can beat the human World Cup champion application of rescue activities for any kind soccer team by 2050. It's obvious that of disaster; and (3) RoboCupJunior, an international building a robot to play a soccer game is an education-based initiative designed to immense challenge; readers might therefore introduce young students to robotics. It is our intention to use since 1997 and showed its epoch-making new RoboCup as a vehicle to promote robotics standard for future RoboCups. One thousand and AI research by offering a publicly appealing four team members from 188 teams from 30 but formidable challenge (Asada et nations around the world participated. It included al. 1999; Kitano et al. 1997). The humanoid league is a big challenge knowledge, this was the largest robotic event with a long-term, high-impact goal, which in history.