Engenharia de Robôs
URI permanente desta comunidadehttps://repositorio.fei.edu.br/handle/FEI/339
Navegar
4 resultados
Resultados da Pesquisa
- Heuristically-accelerated reinforcement learning: A comparative analysis of performance(2014) MARTINS, M. F.; Reinaldo BianchiThis paper presents a comparative analysis of three Reinforcement Learning algorithms (Q-learning, Q(λ)-learning and QS-learning) and their heuristically-accelerated variants (HAQL, HAQ(λ) and HAQS) where heuristics bias action selection, thus speeding up the learning. The experiments were performed in a simulated robot soccer environment which reproduces the conditions of a real competition league environment. The results clearly demonstrate that the use of heuristics substantially improves the performance of the learning algorithms. © 2014 Springer-Verlag.
- Comparison and analysis of the DVG+A∗ and rapidly-exploring random trees path-planners for the robocup-small size league(2019-10-23) DA SILVA COSTA, L.; Flavio Tonidandel© 2019 IEEE.This paper provides an experimental analysis between Dynamic Visibility Graph A Star (DVG+A*) and Rapidly-exploring Random Trees (RRT) path-planners, in order to compare which one is more adequate to the scenario presented in the Small Size League (SSL). The metrics used to compare each algorithm were established based on the characteristics of a SSL game, which demand a short path, low computational cost and a safe distance from the opponent robots. For the comparison, both algorithms were tested in static and dynamic environments. After all the tests, DVG+A∗ has shown the best results.
- Performing and blocking passes in small size league(2019-10-25) LAUREANO, M. A. P.; Flavio Tonidandel© 2019 IEEE.The changes in the Small Size League rules, like increasing the field size and inclusion of more robots in the game have brought greater possibilities of playing and strategies. With the increased complexity of soccer matches, the positioning of the robots has become important as the defense and attack mechanisms. The learning of opposing team game playing has been shown to be effective, but an SSL soccer match indicates the need for solutions that analyze the strategy of the opposing team during the game and make necessary adaptations. This paper proposes the use of the Particle Swarm Optimization (PSO) algorithm as an option to determine the positioning for making and blocking passes during the match. A prototype has been developed to validate the configuration parameters. Experiments in a simulator and analysis of game logs have demonstrated the feasibility of applying the PSO algorithm to find the robots positions.
- Qualitative case-based reasoning and learning(2020-03-20) HOMEM, THIAGO PEDRO DONADON; Paulo Santos; COSTA, ANNA HELENA REALI; Reinaldo Bianchi; MANTARAS, RAMON LOPEZ DEThe development of autonomous agents that perform tasks with the same dexterity as performed by humans is one of the challenges of artificial intelligence and robotics. This motivates the research on intelligent agents, since the agent must choose the best action in a dynamic environment in order to maximise the final score. In this context, the present paper introduces a novel algorithm for Qualitative Case-Based Reasoning and Learning (QCBRL), which is a case-based reasoning system that uses qualitative spatial representations to retrieve and reuse cases by means of relations between objects in the environment. Combined with reinforcement learning, QCBRL allows the agent to learn new qualitative cases at runtime, without assuming a pre-processing step. In order to avoid cases that do not lead to the maximum performance, QCBRL executes case-base maintenance, excluding these cases and obtaining new (more suitable) ones. Experimental evaluation of QCBRL was conducted in a simulated robot-soccer environment, in a real humanoid-robot environment and on simple tasks in two distinct gridworld domains. Results show that QCBRL outperforms traditional RL methods. As a result of running QCBRL in autonomous soccer matches, the robots performed a higher average number of goals than those obtained when using pure numerical models. In the gridworlds considered, the agent was able to learn optimal and safety policies.