Steels, L. and Brooks, R.A., eds (1995) The Artificial Life Route to Artificial Intelligence: Building Embodied Situated Agents, Erlbaum  Home/Search   Document Not in Database   Summary   Related Articles   Check

....this problem, either implicitly or explicitly (e.g. Steels, 1996d; Stein, 1997; Gaussier et al. 1998] For example, in Steels use of his process networks he resists labelling any set of processes as a behaviour. Any process may play a role in a number of difference behaviour systems ([Steels and Brooks, 1995] Chapter 5) Although some proposals for a solution to this problem have been made, we are far from a consensus about how agents can be built to avoid this pitfall. Steels proposes that sets of additional constraints, in the form of recurrent patterns, on how process networks are assembled may ....

....mechanism, no other schemes meet this requirement adequately. Steels Process Networks Steels has recognised that behaviour is a dynamical process. He has proposed his process networks as a method of creating behaviour as a set of processes that link continuously varying quantities ([Steels and Brooks, 1995] chapter 5) There is no need to discretise or categorise their states. Each of the processes may play a role in a number of difference behaviour systems. The behaviour of the systems becomes an emergent property of the interaction between the processes and the environment. Process networks are ....

Steels, Luc and Brooks, Rodney (eds.), " The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents", Lawrence Erlbaum Associates, Publishers, Hove UK. ISBN-0-8058-1518-X, 1995.

....due to the computational distance between them; 3) A system based on such an architecture is not fault tolerant. Therefore, a lot of recent work on robot control aims at finding efficient sensor based solutions to shorten the distance between perception and action. The behaviour based approaches [Se95] use the parallel instead of hierarchical control structure and try to implement the so called embodied, situated behaviours without building complete world models and planning any actions and motions. Obviously, both control architectures have advantages and disadvantages, so it becomes an ....

....robot controller, external and internal sensor data should be applied directly in each control cycle instead of building and updating the world model. If sensor data is coupled with motion control in a simple form, the robot can decide its reaction in time. The word situatedness used by Brooks ([Se95]) contains the similar idea. Simon [Sim69] summarised with bounded rationality the principle that human beings often use only incomplete or imprecise knowledge for problem solving. Sensor data needed for direct integration in motion control possess the following features: ffl They are ....

L. Steels and R. Brooks (editors). The Artificial Life Route to Artificial Intelligence: building Embodied, Situated Agents. Lawrance Erlbaum Associates Publishers, 1995.

.... System to be created and so generate new behavior patterns from the interaction of its basic components (emergence) Based on this, in the last years it has came up a new approach to conceive intelligent systems called the Bottomup AI or better known as Behavior based Artificial Intelligence [14, 3] which is concerned with designing artificial systems to exhibit intelligent behavior rather than giving to them knowledge representation or reasoning capabilities as in the alternative fashion. In addition, such a behavior requires an strong coupling between the agent s working acting and its ....

Steels, L. & Brooks R. The Artificial Life Route to Artificial Intelligence. Lawrence Erlbaum Associates, UK, 1995.

....external and internal sensor data should be applied directly in each control cycle instead of being used for building and updating the world model. If sensor data is coupled with motion control in a simple form, the robot can determine its reaction in time. The word situatedness used by Brooks [9] develops a similar idea. Simon [8] summarised with the concept of bounded rationality the principle that human beings often use only incomplete or imprecise knowledge for problem solving. Sensor data needed for direct integration into motion control possess the following features: ....

L. Steels and R. Brooks (editors). The Artificial Life Route to Artificial Intelligence: building Embodied, Situated Agents. Lawrance Erlbaum Associates Publishers, 1995.

....from interactions between systems, to SIA research. Two AL research directions are outlined: a) the quest for a logic of life, opposed to b) studying the natural form of complexity in artificial media by constructing systems. The latter is the general research agenda of Embodied AL ( Mae90] SB95] Its starting points for constructing systems are the physical properties of the matter, not abstract formalisms. I interpret concepts like believability , stories and social understanding within this framework of creating life like artifacts. EAL might thus provide a valuable framework ....

....AL seems to be in danger to end up in the same trap, by assuming that we can make a system alive by studying logic and implementing it on computer or robot hardware. Therefore Embodied AL is an alternative approach towards life like artifacts which originated in work e.g. described in [Mae90] and [SB95] The EAL framework towards intelligence and aliveness focuses on the grounding of these conceptions in the concrete embodied, situated system, and studies how (from an observer point of view) interactions with the environment can bias the attribution of intelligence and liveliness . EAL takes ....

Luc Steels and Rodney Brooks, editors. The artificial life route to artificial intelligence: building embodied, situated agents. Lawrence Erlbaum Assoc., 1995.

.... and the mechanisms of multiple situation based actions at a micro level (the one of the agents) A system s behavior can be considered emergent if it can only be specified using descriptive categories which are not necessary to describe the behavior of the constituent components [ 9 ] Steels in [ 27 ] adds that an emergent behavior leads to emergent functionality if the behavior contributes to the system s self preservation and if the system can build further upon it. 2.1 Autonomy In Autonomous Agents, two types of autonomy are commonly pointed out [ 30 ] operational autonomy and behavioral ....

L. Steels and R. Brooks. The artificial life route to artificial intelligence. Building Embodied, Situated Agents. Lawrence Erlbaum, Hillsdale, NJ, USA, 1994.

....to meet the challenges posed by mobile robots. Brooks decomposes a control system into a set of interacting behavior modules described in LISP like programming languages [2] Others use production rules approach as building blocks for the control system [6] 9] or general dynamical systems [25]. Some researchers propose to use artificial neural networks due to its generally smoother search space and its working with very low primitives avoiding using preconceptions about the properties of the systems[11] 13] 1] The use of complex building blocks for the control system of the robot is ....

Luc Steels and Rodney Brooks, editors. The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents. Lawrence Erlbaum Assoc, 1995.

....in a complex and dynamic environment in order to achieve its autonomy. This is one example of the nouvelle AI and Artificial Life research the aim of which is to build autonomous system which can adapt to the world it is embedded in, this world being changing and often unpredictable [13] [20]. To answer this new domain of robotics, Brooks has proposed a new designing methodology for mobile robots called subsumption architecture [2] His approach consists to build behavior based robots. To deal with the autonomous arbitration of the multiple behaviors, Brooks outlines an evolutionary ....

.... its generally smoother search space and its working with very low primitives avoiding using preconceptions about the properties of the systems [14] Finally Luc Steels proposes to use a dynamical systems, called a process network, inspired by the couple map latticed to control the robot behaviors [20]. Boolean Function Controller. In our approach, we assume that the robot behavior is described by a boolean function. It is more simple than the set of behavior modules implemented by augmented finite state machines, although the boolean function represents in an implicit way all the knowledge to ....

Luc Steels and Rodney Brooks, editors. The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents. Lawrence Erlbaum Assoc, 1995. This article was processed using the L a T E X macro package with LLNCS style

....detail that can be utilized automatically both for creating realistic visualizations and functional control code. CAISR Technical Report #TR96 109 Submitted to IEEE ICRA 97 3. 2 Reactive Software Agents for Conveyor Control Our second key concept is the use of reactive software agents [9] to execute conveyor control. Our approach can be explained by analogy as follows. Consider a taxi, a passenger, roads, lift bridges, and a passenger s origin and destination. The taxi represents a pallet; the passenger represents material to be transported by a pallet; lift bridges represents ....

L. Steels and R. Brooks. The Artificial Life Route to Artificial Intelligence. Building Situated Embodied Agents. New Haven: Lawrence Erlbaum Assoc., 1994.

....in a complex and dynamic environment in order to achieve its autonomy. This is one example of the nouvelle AI and Artificial Life research the aim of which is to build autonomous system which can adapt to the world it is embedded in, this world being changing and often unpredictable [20] [29]. To answer this new domain of robotics, Brooks has proposed a new designing methodology for mobile robots called subsumption architecture [3] His approach consists to build behavior based robots. To deal with the autonomous arbitration of the multiple behaviors, Brooks outlines an evolutionary ....

.... with very low primitives avoiding using preconceptions about the properties of the systems [22] 15] Finally Luc Steels proposes to use a dynamical systems, called a process network, inspired by the couple map latticed developed by Kaneko and various coworkers to control the robot behaviors [29][17] 3.1 Boolean Function Controller. In our approach, we assume that the robot behavior is described by a boolean function. It is more simple than the set of behavior modules implemented by augmented finite state machines, although the boolean function represents in an implicit way all the ....

Luc Steels and Rodney Brooks, editors. The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents. Lawrence Erlbaum Assoc, 1995.

....long due to the computational burden; c) A system based on such an architecture is not fault tolerant. Therefore, a lot of recent work on robot control aims at finding efficient sensor based solutions to reduce the temporal delay between perception and action. Brooks subsumption architecture [2, 11] essentially consisted of combining a set of parallel reactive behaviours without building complete world models. The main problems with this architecture are: a) Task directed symbolic goals are difficult to be integrated in the behaviours (thus only insect like intelligence can be emulated) ....

....applied directly in each control cycle instead of building and updating the world model. If sensor data is coupled with motion control in a simple form, the robot can decide its reaction in time. The idea of situatedness 1 More behaviours for a eye head system can be found in [3] by Brooks [11] is comparable to this concept. By bounded rationality Simon [10] summarised the principle that humans often use only incomplete or imprecise knowledge for problem solving. Sensor data needed for direct integration in robot control possess the following properties: They are relative. These ....

L. Steels and R. Brooks (editors). The Artificial Life Route to Artificial Intelligence: building Embodied, Situated Agents. Lawrance Erlbaum Associates Publishers, 1995.

.... robot task in unknown and dynamic environments, researchers have applied evolution based learning algorithms [22] to low level control architecture such as LISP like programming languages [19] 4] 26] finite state automata [32] production rules (classifier systems) 35] 6] 9] process network [28] and neural networks [7] 2] 24] 5] 10] 13] 12] In our experiments, the learning task consists to find the function F , mapping 256 inputs (world states) to 8 outputs (motions) in a search space of 8 256 functions from a given set of observable input output pairs and a reward signal. For ....

Luc Steels and Rodney Brooks, editors. The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents. Lawrence Erlbaum Assoc, 1995.

....of multiple competences and situation based learning at a micro level (the one of the agents) A system s behavior can be considered emergent if it can only be specified using descriptive categories which are not necessary to describe the behavior of the constituent components. Steels in [64] adds that an emergent behavior leads to emergent functionality if the behavior contributes to the system s self preservation and if the system can build further upon it. Numerous works deal with interactions and emergence: some tackle the problem of coordinating interactions [43] 25] 8] 6] ....

L. Steels and R. Brooks. The artificial life route to artificial intelligence. Building Embodied, Situated Agents. Lawrence Erlbaum, Hillsdale, NJ, USA, 1994.

....of ethology, neuroscience, and evolutionary theory to name a few. Looking to animals to glean insight into how to build intelligent autonomous systems, the field has brought the issues of embodiment, situatedness, emergence, task based decomposition, and environmental complexity to the foreground (Brooks Steels 1995). Indeed, the treatment of these issues characterizes various biologically inspired approaches, several of which have been successfully applied to the control of autonomous robots behavior based systems (Brooks 1986) evolution based systems (Cliff et al. 1993) and connectionist systems (Beer ....

Brooks, R., and Steels, L. 1995. The Artificial Life Route to Artificial Intelligence. Hillsdale, New Jersey: Lawrence Erlbaum Associates, publishers.

.... Cliff 1996) to low level control architecture such as LISP like programming languages (Koza 1992) Brooks 1992) Reynolds 1994) finite state automata (Thompson 1995) production rules (classifier systems) Wilson 1987) Dorigo Colombetti 1994) Grefenstette Schultz 1994) process network (Steels Brooks 1995) and neural networks (Floreano Mondada 1996) Beer Gallagher 1992) Parisi, Nolfi, Cecconi 1992) Husbands et al. 1995) Hoshino, Mitsumoto, Nagano 1998) In our experiments, the learning task consists to find the function F , mapping 256 inputs (world states) to 8 outputs (motions) in a ....

Steels, L., and Brooks, R., eds. 1995. The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents. Lawrence Erlbaum Assoc.

....the system) and situation based interactions at a micro level (the one of every agent) Still in [11] a system s behavior can be considered emergent if it can only be specified using descriptive categories which are not necessary to describe the behavior of the constituent components. Steels, in [27], adds that an emergent behavior leads to emergent functionality if the behavior contributes to the system s self preservation and if the system can build further upon it. In [21] emergence is defined by resorting to self organization: self organized dynamic systems composed of a large number of ....

L. Steels and R. Brooks. The artificial life route to artificial intelligence. Building Embodied, Situated Agents. Lawrence Erlbaum, Hillsdale, NJ, USA, 1994.

.... and results of animat research have been extensively reviewed elsewhere, the corresponding papers have mostly emphasized their similarities and differences with respect to those of traditional Artificial Intelligence (Meyer and Guillot, 1991, Wilson, 1991; Maes, 1993; Meyer and Guillot,1994; Steels and Brooks, 1995; Meyer, 1995, 1996 ) Within the framework of such comparisons, the biological roots of the animat approach haven t been granted special attention. This paper will be centered on these roots and will argue that adaptive behaviors generated by biomimetic mechanisms are promising solutions to ....

Steels, L. and Brooks, R. (Eds.) (1995).The Artificial Life Route to Artificial Intelligence.

.... the behavior of people is coordinated with the external world) without the mediation of mental planning (i.e. without explicit inferencing over descriptive 29 models of the world and human behavior) This view of cognition also extends directly to alternative robotic architectures in AI (cf. Steels Brooks, 1995; Clancey, 1997) Vera Simon (1993) claim that the Situated Action approach is not very different from, and could be easily incorporated within, the traditional framework (i.e. the PSS hypothesis) Both approaches, they claim, are based on the same basic architecture, in the center of which ....

Steels, L. & Brooks, R. (Eds.). (1995). The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents. Hillsdale: Lawrence Erlbaum.

....own movement, the agent generates correlations in the interaction with the environment. The second important aspect of sensory motor coordination is the generation of cross modal associations, including proprioceptive cues originating from the motor system (Thelen and Smith, 1984; Scheier and Lambrinos, 1996). Additional support for the principle of sensory motor coordination comes from developmental studies. There is a lot of evidence that concept formation in human infants is directly based on sensory motor coordination (Thelen and Smith, 1984; Smith and Thelen, 1993; see figure 6) The concepts of ....

....basis in introspection. The latter has long ago been demonstrated to be a poor guide to research (Nisbett and Wilson, 1977) References supporting this principle include: Ballard, 1991; Dewey, 1896; Douglas, 1993; Edelman, 1987; Thelen and Smith, 1994; Smith and Thelen, 1993; Scheier and Lambrinos, 1996; Pfeifer and Scheier, in press; Scheier and Pfeifer, 1995; Principle 6: The principle of ecological balance The principle of ecological balance states that there has to be a match between the complexity of the sensors, the actuators, and the neural substrate. Moreover, it states that the ....

In L. Steels and R. Brooks (eds.). The Artificial Life Route to Artificial Intelligence. Hillsdale, N.J.: Erlbaum, 237-263. Scheier, C., and Lambrinos, D. (1996). Categorization in a real-world agent using haptic exploration and active vision. Proc. SAB'96.

....( physically realized ) nature of natural cognition as an alternative basis for understanding and replicating mental phenomena. In so doing, these cognitive scientists have also rediscovered the work of the cyberneticists. See, for example, Varela and Borgine 1992) Port and van Gelder 1995) (Steels and Brooks 1995). This argument has been most extreme and the transition away from the traditional computational metaphor most clear cut in the field of robotics. As articulated by Brooks (1986) the idea is that physical control of a robot is best achieved through a horizontal decomposition in which each ....

Steels, Luc and Rodney Brooks, editors. 1995. The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents. Hillsdale, New Jersey: Lawrence Erlbaum Associates, Inc.

No context found.

Steels, L. and Brooks, R.A., eds (1995) The Artificial Life Route to Artificial Intelligence: Building Embodied Situated Agents, Erlbaum

....is focused on the bottom up approach. i.e. the research on how behaviour can be build up by an agent from its low level (sensory motor) control in such a way that the agent autonomously increase its cognitive capabilities. For an introduction to the behaviour oriented approach of AI see e.g. [38]. There are several methods used in the study of this subject: behaviour oriented architectures, neural network methods or genetic algorithms. Our approach is based on the behaviouroriented architectures. Agents directly connect their sensors to motors in order to control behaviour. Architectures ....

....in a C implementation of PDL, which was first developed in Lisp. This program was developed because of the need for a simple programmable device that controls dynamic behaviour of intelligent autonomous agents [37] The system is born inside the behaviour based approach to autonomous systems [38] and it is designed to be capable of real time interaction between dynamic processes inside the agent and the physical processes of the environment [52] The PDL has its own structure and syntax, although its implementation is embedded in C, thus the C syntax is valid as well. The software ....

[Article contains additional citation context not shown here]

Steels, L. (1994a) Artificial Life route to Artificial Intelligence. In: Artificial Life Journal 1(1), pp 89-125

No context found.

Steels, L. and R. Brooks (1995) The Artificial Life Route to Artificial Intelligence. Building Embodied Situated Agents. Lawrence Erlbaum, New Haven.

No context found.

Steels, L. & Brooks, R., eds (1995), The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents, Lawrence Erlbaum Associates, Hillsdale, New Jersey.

No context found.

Steels, L., and Brooks, R., Eds. "The Artificial Life Route to Artificial Intelligence ", Building Situated Embodied Mechanisms (New Haven, 1995), Lawrence Erlbaum.

No context found.

Steels, L. and Brooks, R. (1995) The Artificial Life Route to Artificial Intelligence. Lawrence Erlbaum, Hillsdale, NJ.

No context found.

L. Steels and R. Brooks, editors. The Artificial Life Route to Artificial Intelligence: Building Embodied, Situated Agents. Hillsdale: Lawrence Erlbaum, 1995.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC