Content uploaded by K. Broda
Author content
All content in this area was uploaded by K. Broda on Jul 07, 2015
Content may be subject to copyright.
A preview of the PDF is not available
Constructing Teleo-reactive Robot Programs.
... Leaving aside the field of learning, we should draw attention to the work of K. Broda's team at Imperial College (London). Taking as base the formal framework for synthesizing Teleo-Reactive robot control programs presented in Broda (2000), they have carried out a series of studies for the design and assessment of Teleo-Reactive policies in multi-agent environments based on analysis and manipulation of situation graphs. The intersections in these graphs represent the possible actions that may be taken from a given node, which in turn represents the environment and the agent's perception of the environment. ...
... The authors stated that the robot navigated successfully around newly discovered obstacles to different goal locations within the experimental environment. The work of Krysia Broda's Group at Imperial College London (Broda 2000;Broda andHogger 2004a, 2010) has proven highly useful for the development of individual or collaborative multiagent (and not necessarily mobile) systems, offering frameworks for policy creation, means of assessing these and simulations of the systems they have designed. One of their most interesting products in this field is the SAGE (Sense, Abduce, Gather, Execute) environment control and monitoring system presented in Broda et al. (2009), which provides a flexible, distributed, open and componentbased approach to environmental monitoring and control. ...
... The authors stated that the robot navigated successfully around newly discovered obstacles to different goal locations within the experimental environment. The work of Krysia Broda's Group at Imperial College London (Broda 2000;Broda andHogger 2004a, 2010) has proven highly useful for the development of individual or collaborative multiagent (and not necessarily mobile) systems, offering frameworks for policy creation, means of assessing these and simulations of the systems they have designed. One of their most interesting products in this field is the SAGE (Sense, Abduce, Gather, Execute) environment control and monitoring system presented in Broda et al. (2009), which provides a flexible, distributed, open and componentbased approach to environmental monitoring and control. ...
The Teleo-Reactive approach designed by N. J. Nilsson offers a high-level programming model for the development of reactive systems such as robotic vehicles. Teleo-Reactive programs are written in a way that allows engineers to define behaviour taking account of goals and changes in the state of the environment. Since Nilsson’s original definition, published in 1994, various researchers have used the Teleo-Reactive paradigm, either applied to a particular domain or extended by adding more capabilities to the original definition. This article provides a systematic literature review of 53 previous Teleo-Reactive-based studies in journals, conference proceedings and the like. The aim of this paper is to identify, appraise, select and synthesize all this high-quality research evidence relating to the use of the Teleo-Reactive paradigm. The literature has been systematically reviewed to offer an overview of the present state of this field of study and identify the principal results that have been obtained thanks to the Teleo-Reactive approach. Finally, this article details the challenges and difficulties that have to be overcome to ensure further advances in the use of this technique.
... The core concepts in our approach, first proposed in [2], are as follows. The world in which the robot operates has a total set O of possible states. ...
... (a) states and perceptions o P (o) 1 [1,1] {a, b} 2 [1] {d, e} 3 [2] {a, c} ...
... The former associates no positional data to the robot or the towers, and so uses the same information about the problem as the predictor. Thus, if the robot picks a block from the 2-tower in the state [1,2], it is indeterminate as to which tower it will see afterwards in the new state [1,1]. By contrast, the positional mode assigns discrete grid coordinates to the robot and the towers, and exploits these to effect transitions in a manner somewhat closer to physical reality through knowing precisely where the robot is located and what it is seeing. ...
This paper presents a new method for predicting the values of policies for cloned multiple teleo-reactive robots operating in the con- text of exogenous events. A teleo-reactive robot behaves autonomously under the control of a policy and is pre-disposed by that policy to achieve some goal. Our approach plans for a set of conjoint robots by focusing upon one representative of them. Simulation results reported here in- dicate that our method aords a good degree of predictive power and scalability.
... Its goal-orientedness is thus explicit in the policy. By contrast, we make two key assumptions about the kind of TR-agents studied in this paper, and first introduced in [3]; namely that they have (i) little or no access to cognitive resources, such as beliefs or reasoning systems, and (ii) only partial observational capability, in that their perceptions may not fully capture the whole environmental state, whether a goal state or otherwise. In particular, a policy contains no rule specifically associated with a goal. ...
... In this case other can be identified with whatever external source is responsible for the exogeneity. Our use of such representative entities, which we first investigated in [3], is somewhat similar in motivation to [10] in which the focus is upon MDP-based design, and to [16] in which the focus is upon design by reinforcement learning. We have so far applied it only to cloned agents, that is, ones all having the same policy. ...
... Thefigure also shows, for each perception p, the set O(p) of associated states and the set A(p) of associated actions. o ∈ O 1 [1, 1, 1] 2 [1, 2] 3 [3] 4 [1, 1] 5 [2] p ∈ P O(p) A(p) a s0, h {4, 5} {l, w} b s1, h {4} {l, w} c s2, h {5} {l, w} d s0, nh {1, 2, 3} {w} e s1, nh {1, 2} {k, w} f s2, nh {2} {k, w} g s3, nh {3} {k, w}Figure 2 shows the resulting unrestricted situation graph G. For the sake of compactness, a situation such as (2, f ) is shown there simply as 2f . ...
A policy for a minimal reactive agent is a set of condition-action rules used to determine its response to perceived environmental stimuli. When the policy pre-disposes the agent to achieving a stipulated goal we call it a teleo-reactive policy. This paper presents a framework for constructing and evaluating teleo-reactive policies for one or more minimal agents, based upon discounted-reward evaluation of policy-restricted subgraphs of complete situation-graphs. The main feature of the method is that it exploits explicit and definite associations of the agent's perceptions with states. The combinatorial burden that would potentially ensue from such associations can be ameliorated by suitable use of abstractions. The framework allows one to plan for a number of agents by focusing upon the behaviour of a single representative of them. It allows for varied behaviour to be modelled, including communication between agents. Simulation results presented here indicate that the method affords a good degree of scalability and predictive power.
... The great advantage of the approach is its ability to react robustly to these changes, due to the continuous computation of the conditions of the rules. In short, the approach provides a useful mechanism to system developers for the goal-oriented specification of reactive systems (see [Rajan 2010], [Gubisch 2008] and [Broda 2000]). ...
... In the same line, [Hawthorne 2011] analyzes the implications that may arise when a rule is incorrectly conceived for a given TR objective, showing how to organize the rules within a goal. [Broda 2000] provides a formal method of systematically obtaining TR programs through a constructive process based on graphs, in which the nodes represent pairs [perception, observable real world] and the transitions are labeled with the actions the robot can perform in each state. This proposal, although perfectly focused from the point of view of building TR programs, like other contributions oriented to Software Engineering, has the disadvantage that the space of states is potentially huge. ...
Context: the Teleo-Reactive (TR) approach offers many possibilities for goal-oriented modeling of reactive systems, but it also has drawbacks when the number of interactions among agents is high, leading to barely legible specifications and losing the original benefits of the approach.
Objective: this work combines the TR paradigm with statecharts and provides advantages for modeling reactive systems and removing the shortcomings detected.
Method: a basic example is adopted to reveal the problem that appears when agents are modeled only with the TR approach and have frequent interactions with others. This paper proposes an extension to the TR approach that integrates the modeling using statecharts. A transformation procedure from statecharts to TR programs makes it possible to continue using the infrastructure of existing execution platforms such as TeleoR. The approach has been validated for a particular domain by considering a more complex case study in which traditionally there have been no results on the application of the TR paradigm. A survey was carried out on students to verify the benefits of the approach.
Results: a method to consider statecharts when modeling TR programs.
Conclusions: statecharts can facilitate the adoption of the TR approach.
... The former associates no positional data to the agent or the towers, and so precisely mirrors the information content of the original problem formulation. This means, for instance, that when the agent picks a block from the 2-tower in the state [1,1,2], it is indeterminate as to which tower it will see afterwards in the new state [1, 1, 1]. By contrast, the positional mode assigns discrete grid coordinates to the agent and the towers, and exploits these to effect transitions in a manner somewhat closer to physical reality through knowing precisely where the agent is located and what it sees. ...
... Related work on agent planning by learning is given in [9] and [12]. The formulation of teleo-reactive programs as restrictions of situation graphs was introduced in [2]. The discounted reward principle [7] has recently been used to plan cooperative and communicative teleo-reactive agents [3,4]. ...
A method is presented for designing an individual teleo- reactive agent, based upon discounted-reward evaluation of policy-restricted subgraphs of complete situation-graphs. The main feature of the method is that it exploits explicit and denite associations of the agent's percep- tions with states. The combinatorial burden that would potentially ensue from such associations can be ameliorated by suitable use of abstraction. For various BlocksWorld examples, the agent's predicted behaviour is compared with simulation results to provide insight into the method's power and scalability. Teleo-reactive agents (10) react to their perceptions of the world by obeying an internal program (or policy) mapping perceptions to actions. The simplest policy structure is a set of mutually-exclusive production rules of the form perception ! action, usually intended to control durative behaviour: given some current perception the agent performs the corresponding action until acquiring a new perception, whereupon it reacts likewise to that. Such an agent may or may not have sucien t perceptive capability to know, at any instant, the entire state of the world. An agent of the kind described in (11) is presumed at least capable of perceiving an intended goal state whenever that state arises, and is accordingly designed with that capability in mind. Its program includes an explicit test for the goal state, whilst the nature and ordering of its rules are inferred by reductive analysis of that test. Its goal-orientedness is thus explicit in the program. A teleo-reactive agent of the kind studied in this paper is, by contrast, pre- sumed incapable of perceiving the entirety of any state, whether a goal state or otherwise. In particular, its program contains no rule specically associated with a goal. The design process now relies not upon goal-reductive analysis but upon comparing the extents to which alternative programs dispose the agent towards achieving a goal { as judged, for instance, by a discounted-reward principle. A program identied on this basis is implicitly goal-oriented. Our design process is based upon graphs that relate combinations of per- ception and objective state, and so diers from approaches that employ graphs relating perceptions alone. These typically estimate the distribution of states
... The main advantage of this approach is its robustness, due to the continuous computation of the conditions on which agent actions are based. In short, the TR paradigm offers a useful approach for developing systems when a goal-oriented specification is employed (see [4][5] [6] for further details of this approach). ...
AbstractContext TeleoR is an extension and implementation of Teleo-Reactive (TR) language for defining the behavior of reactive systems when the consideration of timing constraints is a matter of interest. Objective This paper analyzes how to consider real-time constraints when a TR approach is followed from modeling to implementation. Method After carrying out a study of the type of timing constraints from the TR perspective, the possibility of using TeleoR for incorporating such constraints was considered. Some extensions on TRiStar notation were then made to represent temporal requirements. A drone-based case study was carried out to demonstrate the usefulness of this approach. Finally, a survey was conducted to validate the approach. Results TeleoR can, to a great extent, support the kind of real-time constraints required for developing real-time systems, offering a direct solution to five of the eight temporal requirements identified, which can be implemented using the basic features of the language. Conclusions Considering real-time requirements should be part of the specification of reactive systems implemented when using the TR approach and should be supported by the implementation platform. In this regard, TeleoR offers reasonable possibilities that should be extended by taking into account the limitations identified here.
... One representative instance of reactive systems is the mobile robotics domain, where the TR paradigm has provided very interesting results [4][5] [6], since TR-based systems inherently recover from errors and unexpected events whilst proceeding towards their goals. This domain embraces applications in which robotic vehicles work autonomously in non-structured environments. ...
Context: Teleo-Reactive (TR) specifications allow engineers to define the behavior of reactive systems while taking into account goals and changes in the state of the environment.
Objective: This article evaluates two different Goal Oriented Requirements Engineering notations, i* and KAOS, to determine their understandability level for specifying TR systems.
Method: A controlled experiment was performed by two groups of Bachelor students. Each group first analyzed a requirements model of a mobile robotic system, specified using one of the evaluated languages, and then they filled in a questionnaire to evaluate its understandability. Afterwards, each group proceeded similarly with the model of another system specified with the second language.
Results: The statistical analysis of the data obtained by means of the experiment showed that the understandability of i* is higher than that of KAOS when modeling TR systems.
Conclusion: Both languages are suitable for specifying TR systems although their notations should be specialized to maximize the understandability attribute. i* surpasses KAOS due to two main reasons: i* models represent dependencies between agents and goals or tasks; and notational differences between tasks and goals in i* are more evident than those between goals and requirements in KAOS.
... The analysis yields an overall policy value that will depend upon, inter alia, the designated goal situation(s) and whatever rewards and probabilities are assigned to the graph's arcs by the designer. The method is called the situation-graph-framework (SGF) to reflect its reliance upon explicit situation graphs, and was first reported in [1]. The SGF framework can be distinguished from both the standard MDP and POMDP frameworks [7, 4]. ...
Abstraction is a valuable tool for dealing with scalability in large state space contexts. This paper addresses the design, using abst raction, of good poli- cies for minimal autonomous agents applied within a situation-graph-framework. In this framework an agent's policy is some function that map s perceptual inputs to actions deterministically. A good policy disposes the agent towards achieving one or more designated goal situations, and the design process aims to identify such policies. The agents to which the framework applies are assumed to have only partial observability, and in particular may not be abl e to perceive fully a goal situation. A further assumption is that the environmen t may influence an agent's situation by unpredictable exogenous events, so th at a policy cannot take advantage, of a reliable history of previous actions. The Be llman discount mea- sure provides a means of evaluating situations and hence the overall value of a policy. When abstraction is used, the accuracy of the method can be significantly improved by modifying the standard Bellman equations. This paper describes the modification and demonstrates its power through compari son with simulation results.
... InCoffey and Clark (2006)this last work provided a basis for a hybrid robot control architecture based on agent control schemes (Beliefs–Desires–Intentions, BDI) complete with TR programming. InBroda et al. (2000), a construct process is presented in which TR programs can be constructed systematically. This work is particularly interesting in that it provides methodological steps for the definition of TR programs. ...
This paper describes a method for constructing and evaluating teleo-reactive policies for one or more agents, based upon discounted-reward
evaluation of policy-restricted subgraphs of complete situation-graphs. The combinatorial burden that would potentially ensue
from state-perception associations can be ameliorated by suitable use of abstractions and empirical simulation results indicate
that the method affords a good degree of scalability and predictive power. The paper formally analyses the predictive quality
of two different abstractions, one for applications involving several agents and one for applications with large numbers of
perceptions. Sufficient conditions for reasonable predictive quality are given.
ResearchGate has not been able to resolve any references for this publication.