Figure 7 - uploaded by Rüdiger Dillmann
Content may be subject to copyright.
The mobile system PRIAMOS. To eeciently integrate PRIAMOS' several sensor systems in order to not only reene geometric maps 19] but to generate maps containing topologic and semantic information has proven to be a very diicult task that demands learning capabilities. Therefore, the sophisticated user interface that has been developed for easy teleoperation and task speciication 13] will now be extended to provide examples that are used to let the robot learn exploration strategies. Since these exploration strategies require the coordinated use of the robot's sensor systems and its basic navigation skills (such as collision avoidance skills), this setting provides an immediate application for the multi-agent learning framework described within this paper. Consider the task of programming PRIAMOS to pass through a door. This task comprises activities such as approching the door, identiication of the door's state (opened, closed), and an exact positioning of the robot in front of the door. However, each of these activities, if performed manually, results in a single RCA. Within each R CA, the user operates the mobile platform and the vision system to achieve subgoals that can be automatically identiied. Each o f these subgoals, such a s door-detected, positioned, and through-door is the result of a speciic strategy and therefore located at the end of an RCS. Finally, we obtain for the door-passing task a sequence of subgoals to be achieved, the corresponding system states, and an indication of the agents that are responsible for achieving the subgoals.
Source publication
![Figure 1: The agent-oriented system architecture (from 10]).](profile/Ruediger-Dillmann/publication/3679340/figure/fig3/AS:667767749505038@1536219646179/The-agent-oriented-system-architecture-from-10_Q320.jpg)
While distributed control architectures have many advantages over
centralized ones, such as their inherent modularity and fault tolerance,
a major problem of such architectures is to ensure the goal-oriented
behaviour of the controlled system. This paper presents a framework
within which the coordination skills required for goal-orientedness are
le...
Context in source publication
Context 1
... test bed for experiments using the multi-agent approach will be the mobile robot PRIAMOS 3], Fig. 7. PRIAMOS is a mobile system with three degrees of freedom, i.e. motion in longitudinal and transverse directions and rotation around the centre of the vehi- cle. A set of ultrasonic range sensors is mounted on a circle around the centre of the vehicle. It is also equipped with the active stereo-vision head KASTOR and a laser scanner ...
Similar publications
We present the results of precise static and kinematic point-positioning solutions obtained in post-processing, with a recursive Kalman filter and smoother estimator (i.e., a state space approach). The observations are dual- frequency GPS carrier phase and pseudo-range, treated as two distinct data types, each with its own measurement equations. Th...
Citations
... In multi agent controllers, there are more than one controller or optimization/adaptation criteria that we called them agent and each agent is trying to reach to its goal independently [1, 2]. Each agent analyzes the control results with predefined criteria and produces an emotional feedback. ...
... If control action becomes larger, it causes more stress in the critic output. The output of this critic is a signal between [0, 1] that 0 indicate better controller performance. The fuzzy sets and rules base of this critic is shown infigure 3. ...
In this paper, a multi agent controller for meeting different criteria, based on neuro-fuzzy controller is presented. The proposed controller is motivated by the affective and emotional faculties in human begins, which constantly evaluate the current states with respect to the achievement of the desired goals. For meeting different criteria, the controller consists of several critic agents that each agent tries to meet its goal. The combination of emotions of these agents applies on the controller in order to adapt the learning coefficients to achieve predefined criteria and goals. The proposed controller, also continuously evaluates the current states from critic agents and incremental achievement or disachievement of the set objectives, and self tune its control action accordingly. The controller is based on intelligent neuro-fuzzy architecture that suitable for online training algorithms. The effectiveness of the proposed method is demonstrated trough examples in which the proposed system is used for reducing control effort and tracking error simultaneously. The contribution of critic's emotions in multi criteria satisfaction is highlighted through these examples.
... In recent years, there have been many researches in multi agent controllers [1], [2]. In multi agent controllers, there are more than one controller or optimization/adaptation criteria that we called them agent and each agent is trying to reach to its goal independently. ...
In this paper, we describe a multi agent controller for meeting different criteria, based on emotional learning. Our proposed controller is motivated by the affective and emotional faculties in human begins, which constantly evaluate the current states with respect to the achievement of the desired goals. For meeting different criteria, the controller consists of several critic agents that each agent tries to meet its goal. The combination of emotions of these agents applies on the controller in order to adapt the learning coefficients to achieve predefined criteria and goals. Our proposed controller, also continuously evaluate the current states from critic agents and incremental achievement or disachievement of the set objectives, and self tune its control action accordingly. The controller is based on intelligent neurofuzzy architecture that suitable for online training algorithms. The effectiveness of the total multi agent emotional control system (MEAC) is demonstrated trough examples in which the proposed system is used for reducing control effort and tracking error simultaneously. The contribution of critic's emotions in multi criteria satisfaction is highlighted through these examples.
... One approach is to have the process controlled andlor managed by one dedicated agent called e.g. manager agent (Kaiser et al., 1996b). In other systems the burden of further decomposing a task, 601 building a team for task execution etc. is given to each agent that initially applied for a task and got it assigned (Langle et al., 1995). ...
Currently, an inportant topic in robotic reasearch systems is the design and development of learning Multi-Agent Systems (MAS). One major advantage of these systems is the fact that several agent work towards a common goal, having different specializations for specific subtasks. Especially learning MAS in cooperation with a human teacher seem to be a very promising approach for complex manipulation and production tasks. Since agents can join and leave the system at any time, it is important that knowledge acquired by single agents can be transferred or propagated between agents, to ensure that knowledge is not lost, if agents leave the system. Therefore, techniques will be presented to represent extentable action knowledge for task solutions in an agent's knowledge base and additionally, algorithms for propagating this knowledge between agents efficiently and with minimum communication effort.
... One approach is to have the process controlled and/or managed by one dedicated agent called e.g. manager agent (Kaiser et al., 1996b). In other systems ing a team for task execution etc. is given to each agent that initially applied for a task and got it assigned (LL angle et al., 1995 ). ...
Currently, an important topic of robotic research is the design and development of multi-agent robot systems (MASs). In these a number of autonomous robots cooperate and coordinate themselves in order to pursue given goals. The agents of an MAS not only have to work autonomously or in cooperation with other agents, but in dynamic, relatively unstructured environments. Therefore, the agents require agent-specific but flexible skills to cope with their tasks and the environment's variability. On the other hand, the actions to be performed by agents in an MAS have to meet certain requirements imposed by the MAS's structure. The representation of actions has to support planning, inter-agent communication, task negotiation etc. In this paper, we describe a method of combining the agent-specific nature of skills with the requirements for a general action knowledge representation inherent to MASs, by presenting elementary operations (EOs) that provide an appropriate interface.
... One approach is to have the process controlled and/or managed by one dedicated agent called e.g. manager agent (Kaiser et al., 1996b). In other systems the burden of further decomposing a task, building a team for task execution etc. is given to each a g e n t that initially applied for a task and got it assigned (L angle et al., 1995). ...
Multi-Agent Systems (MAS) do play an important role in the construction of fault tolerant and robust robot systems. One major advantage of MAS is the fact that multiple agents work towards a common goal, having different skills for specific subtasks. Usually, agents have to use a common description of the actions to be carried out. Since agents can join and leave the MAS at any time, it is important that knowledge acquired by single agents can be transferred or propagated between agents, to ensure that knowledge is not lost, in the case that an agent leaves the system. In this paper, techniques will be presented that enable representation of common extendable action knowledge for task solutions in an agent’s knowledge base and additionally, algorithms for propagating this knowledge between agents efficiently and with minimum required communication effort.
This paper proposes a robot control programming system for the specification, execution and estimation of multi-sensor, multi-robot, constraint-based robot applications. A multi-layered set of specification primitives is supported at, respectively, constraint, skill, and task level. These layers provide appropriate abstractions for both application builders and application users. The implementation challenges of the presented approach are discussed. A few examples of real experiments are given that use the proposed specification approach to various degrees.
Multi-agent systems (MASs) do play an important role in the construction of fault tolerant and robust robot systems. One major advantage of MAS is the fact that multiple agents work towards a common goal, having different skills for specific subtasks. Usually, agents have to use a common description of the actions to be carried out. Since agents can join and leave the MAS at any time, it is important that knowledge acquired by single agents can be transferred or propagated between agents, to ensure that knowledge is not lost, in the case that an agent leaves the system. In this paper, techniques will be presented that enable representation of common extendable action knowledge for task solutions in an agent’s knowledge base and additionally, algorithms for propagating this knowledge between agents efficiently and with minimum required communication effort.
In this paper we present a novel approach for coordinating a group of autonomous mobile robots based on behavioral paradigm using finite sate automation (FSA) based reasoning. To achieve coordination these robots do not communicate explicitly with each other but rely on information gathered from the environment. A robot's environment can be modified by its own actions as well as actions of other robots. In our approach a robot senses these modifications through stigmergic sampling of the environment. This sensed information is then used to change the current state of the robot and choose its next action as well. Group level behavior emerges from local behaviors of individual robots and without the robots involving in any active signaling or communication. We claim this approach to be a better alternative than using explicit communication.
In this paper, the multi-agent learning methods in an uncertain environment are addressed. The advantages and disadvantages of each algorithm are given. Rationality and convergence are the two main properties of multi-agent learning algorithms. However, it is very difficult to achieve both properties simultaneously. Minmax-Q learning is guaranteed to converge to equilibrium but there is no guarantee that this is the best response to the actual opponent. Therefore, Minmax-Q is not rational. In contrast, opponent modeling is rational but not convergent. Reinforcement learning using a variable learning rate and simultaneously achieves both properties. To reduce the dimension of state space, modular Q-learning and multilayered reinforcement learning are presented. The presented methods are not exhaustive, but they highlight the major methods used by researchers in the past years.
Despite the numerous advances in human-robot interaction, most development systems still require that users have substantial knowledge of procedural-programming techniques as well as the specific robot system at hand. For the vast majority of the population, this effectively precludes the use of robots in most cases. If robots are to make headway into everyday situations, then users must be able to program robots in a more natural and intuitive manner. This dissertation explores a method of programming robots to automate motor tasks by inferring the intent of users based on demonstrations of a task. In order to understand such a system, we decompose it into simpler components: modeling user subgoal selection and the response of users to different conditions.
