Main Research Thesis Publications C.V.

[1] Martin Stolle and Christopher Atkeson. Transfer of policies based on trajectory libraries. In Proceedings of the International Conference on Intelligent Robots and Systems (IROS 2007), 2007. [ bib | http | .pdf ]
Recently, libraries of trajectory plans have been shown to be a promising way of creating policies for difficult problems. However, often it is not desirable or even possible to create a new library for every task. We present a method for transferring libraries across tasks, which allows us to build libraries by learning from demonstration on one task and apply them to similar tasks. Representing the libraries in a feature-based space is key to supporting transfer. We also search through the library to ensure a complete path to the goal is possible. Results are shown for the Little Dog task. Little Dog is a quadruped robot that has to walk across rough terrain at reasonably fast speeds.

[2] Martin Stolle and Christopher G. Atkeson. Knowledge transfer using local features. In Proceedings of the IEEE Symposium on Approximate Dynamic Programming and Reinforcement Learning (ADPRL 2007), 2007. [ bib | http | .pdf ]
We present a method for reducing the effort required to compute policies for tasks based on solutions to previously solved tasks. The key idea is to use a learned intermediate policy based on local features to create an initial policy for the new task. In order to further improve this initial policy, we developed a form of generalized policy iteration. We achieve a substantial reduction in computation needed to find policies when previous experience is available.

[3] Martin Stolle and Christopher G. Atkeson. Policies based on trajectory libraries. In Proceedings of the International Conference on Robotics and Automation (ICRA 2006), 2006. [ bib | http | .pdf ]
We present a control approach that uses a library of trajectories to establish a global control law or policy. This is an alternative to methods for finding global policies based on value functions using dynamic programming and also to using plans based on a single desired trajectory. Our method has the advantage of providing reasonable policies much faster than dynamic programming can provide an initial policy. It also has the advantage of providing more robust and global policies than following a single desired trajectory. Trajectory libraries can be created for robots with many more degrees of freedom than what dynamic programming can be applied to as well as for robots with dynamic model discontinuities. Results are shown for the “Labyrinth” marble maze, both in simulation as well as a real world version. The marble maze is a difficult task which requires both fast control as well as planning ahead.

[4] Martin Stolle. Automated discovery of options in reinforcement learning. Master's thesis, McGill University, February 2004. [ bib | http | .pdf ]
AI planning benefits greatly from the use of temporally-extended or macro-actions. Macro-actions allow for faster and more efficient planning as well as the reuse of knowledge from previous solutions. In recent years, a significant amount of research has been devoted to incorporating macro-actions in learned controllers, particularly in the context of Reinforcement Learning. One general approach is the use of options (temporally-extended actions) in Reinforcement Learning. While the properties of options are well understood, it is not clear how to find new options automatically. In this thesis we propose two new algorithms for discovering options and compare them to one algorithm from the literature. We also contribute a new algorithm for learning with options which improves on the performance of two widely used learning algorithms. Extensive experiments are used to demonstrate the effectiveness of the proposed algorithms.

[5] Martin Stolle and Doina Precup. Learning options in reinforcement learning. Lecture Notes in Computer Science, 2371:212-223, 2002. [ bib | http | .pdf ]
Temporally extended actions (e.g., macro actions) have proven very useful in speeding up learning, ensuring robustness and building prior knowledge into AI systems. The options framework (Precup, 2000; Sutton, Precup & Singh, 1999) provides a natural way of incorporating such actions into reinforcement learning systems, but leaves open the issue of how good options might be identified. In this paper, we empirically explore a simple approach to creating options. The underlying assumption is that the agent will be asked to perform different goal-achievement tasks in an environment that is otherwise the same over time. Our approach is based on the intuition that “bottleneck” states, i.e. states that are frequently visited on system trajectories, could prove to be useful subgoals (e.g. McGovern & Barto, 2001; Iba, 1989). We present empirical studies of this approach in two gridworld navigation tasks. One of the environments we explored contains bottleneck states, and the algorithm indeed finds these states, as expected. The second environment is an empty gridworld with no obstacles. Although the environment does not contain bottleneck states, our approach still finds useful options, which essentially allow the agent to travel around the environment more quickly.

[6] Francois Rivest, Martin Stolle, and Thomas Shulz. LNSC cascade-correlation simulator applet. WWW, 2001. [ bib | http ]

This file has been generated by bibtex2html 1.88.

Main Research Thesis Publications C.V.
2006-07-10   MADE BY MARTIN