Forward and Bidirectional Planning Based on Reinforcement Learning and Neural Networks in a Simulated Robot (original) (raw)

Planning with a functional neural network architecture

IEEE Transactions on Neural Networks, 1999

This article introduces the concept of planning in an interactive environment between two systems: the challenger and the responder. The responder's task is to produce behavior that relates to the challenger's behavior through some response function. In this setup, we concentrate planning on the responder's actions and use the produced plan in order to control the responder. In general, the responder is assumed to be a nonlinear system whose input-output (I/O) map may be expressed by a Volterra series. The planner uses an estimate of the challenger's future output sequence, the response function, and a model of the responder's I/O relation implemented through a functional artificial neural network (FANN) architecture, in order to produce the input sequence that will be applied to the responder in the future, in parallel-time with the challenger's corresponding output sequence. The responder accepts input from the planner, which may be combined with feedback information, in order to produce an output sequence that relates to the challenger's output sequence according to the response function. The importance of planning for the generation of smooth behavior is discussed, and the effectiveness of the planner's implementation using neural network technology is demonstrated with an example.

Reinforcement Learning and Automated Planning

2008

This article presents a detailed survey on Artificial Intelligent approaches, that combine Reinforcement Learning and Automated Planning. There is a close relationship between those two areas as they both deal with the process of guiding an agent, situated in a dynamic environment, in order to achieve a set of predefined goals. Therefore, it is straightforward to integrate learning and planning, in a single guiding mechanism and there have been many approaches in this direction during the past years. The approaches are organized and presented according to various characteristics, as the used planning mechanism or the reinforcement learning algorithm.

A framework for plan execution in behavior-based robots

Proceedings of the 1998 IEEE International Symposium on Intelligent Control (ISIC) held jointly with IEEE International Symposium on Computational Intelligence in Robotics and Automation (CIRA) Intelligent Systems and Semiotics (ISAS) (Cat. No.98CH36262), 1998

This paper presents a conceptual architecture for autonomous robots that integrates behaviour-based and goal-directed action as by following a traditional action plan. Dual Dynamics is the formalism for describing behaviourbased action. Partial-order propositional plans, which get generated by GRAPHPLAN , are used as a basis for acting goal-directedly; the concept is suitable for using other planning methods and plan representations, though. The paper presents the corresponding action and plan representations at the plan side and at the behaviour side. Moreover, it describes how behaviour-based action is biased towards executing a plan and how information from the behaviour side is fed back to the plan side to determine progress in the plan execution.

Neural network-based modeling of robot action effects in conceptual state space of real world

Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems. IROS '96, 1996

This paper gives the concept of an autonomous robotic agent that is capable of showing both machine learning and reactive behavior. The first methodology is used to collect information about the environment and to plan robot actions based on this information while the robot is performing tasks. Processing and storing information obtained during several task executions is called lifelong learning. Reactive behavior, the second desirable feature of an autonomous robot, is needed to execute the actions in a dynamically changing environment.

Dynamical neural networks for planning and low-level robot control

Systems, Man and Cybernetics, …, 2003

We use dynamical neural networks based on the neural field formalism for the control of a mobile robot. The robot navigates in an open environment and is able to plan a path for reaching a particular goal. We will describe how this dynamical approach may be used by a high level system (planning) for controlling a low level behavior (speed of the robot). We give also results about the control of the orientation of a camera and a robot body.

Reinforcement Learning and Automated Planning: A Survey

2012

Comparing Action as Input and Action as Output in a Reinforcement Learning Task

International Journal of Computer Applications, 2013

Generalization techniques are useful for enabling an agent to be able to approximate the value of states it has not encountered so far in reinforcement learning. They are also useful as memory use minimization mechanisms in situations where the state space is too large such that is infeasible to represent every state in the state space in the computer memory. Artificial Neural Networks are one generalization technique that is usually employed. Various network structures have been proposed in literature. In this study, two of the structures that have been proposed were implemented in a robot navigation task and their performance compared. The results indicate that having a network structure where there is an output node for each of the possible actions, is superior to the structure in which the selected action is fed as an input to the network and its value output by the single network output node.

Planning in Artificial Intelligence

A Guided Tour of Artificial Intelligence Research, 2020

In this chapter, we propose a non-exhaustive review of past works of the AI community on classical planning and planning under uncertainty. We first present the classical propositional STRIPS planning language. Its extensions, based on the problem description language PDDL have become a standard in the community. We briefly deal with the structural analysis of planning problems, which has initiated the development of efficient planning algorithms and associated planners. Then, we describe the Markov Decision Processes framework (MDP), initially proposed in the Operations Research community before the AI community adopted it as a framework for planning under uncertainty. Eventually, we will describe innovative (approximate or exact) MDP solution algorithms as well as recent progresses in AI in terms of knowledge representation (logics, Bayesian networks) which have been used to increase the power of expression of the MDP framework. R. Sabbadin (B)

Logical sensor/actuator : knowledge-based robotic plan execution

Journal of Experimental & Theoretical Artificial Intelligence, 1997

Complex tasks are usually described as high-level goals, leaving out the details on how to achieve them. However, to control a robot, the task must be described in terms of primitive commands for the robot. Having the robot move itself to and through an unknown, and possibly narrow, doorway is an example of such a task. We show how the transformation from high-level goals to primitive commands can be performed at execution time and we propose an architecture based on recon gurable objects that contain domain knowledge and knowledge about the sensors and actuators available. We illustrate our approach using actual data from a real robot.

Forward and Bidirectional Planning Based on Reinforcement Learning and Neural Networks in a Simulated Robot (original) (raw)

Related papers