Sahota, Michael K. (1994), Action Selection for Robots in Dynamic En- vironments through Inter-behaviour Bidding, Proceedings of the Third International Conference on Simulation of Adaptive Behavior (SAB-9).  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... an algorithmic implementation (see [Tyrrell, 1993] for a survey, and for many diculties in translating the conceptual models into computational ones) The action selection models that do lend themselves to algorithmic implementation (e.g. see [Brooks, 1991, Rosenblatt, 1995, Blumberg, 1994, Sahota, 1994, Aylett, 1995] then typically require a consid erable design effort. In the literature, one sees formulas taking weighted sums of various quantities in an attempt to estimate the utility of actions. There is much hand coding and tuning of parameters (e.g. see [Tyrrell, 1993, 9] until the ....

....structured in any hierarchy and are not even necessarily cooperative. In the model I introduce in 5, an agent will control the entire body on its own if allowed. It is generally frustrated by the presence of other agents. Baum also uses agent in his Economy of Mind [Baum, 1996] Behavior (e.g. [Sahota, 1994, Mataric, 1994] is probably the term in most common use, but is often used for modules that are not autonomous actors. For somewhat autonomous, somewhat competing modules within a single body, Brooks, 1986] uses layer (though [Brooks, also uses process) Blumberg, 1994] uses activity (as does ....

[Article contains additional citation context not shown here]

Sahota, Michael K. (1994), Action Selection for Robots in Dynamic En- vironments through Inter-behaviour Bidding, Proceedings of the Third International Conference on Simulation of Adaptive Behavior (SAB-9).

.... are not detailed enough to specify an algorithmic implementation (see [Tyrrell, 1993] for a survey, and for some difficulties in translating the conceptual models into computational ones) The models that do lend themselves to algorithmic implementation (e.g. see [Brooks, 1991, Blumberg, 1994, Sahota, 1994, Aylett, 1995] then typically require a considerable design effort. In the literature, one sees formulas taking weighted sums of various quantities in an attempt to estimate the utility of actions. There is much hand coding and tuning of parameters (e.g. see [Tyrrell, 1993, Ch.9] Aylett, ....

Sahota, Michael K. (1994), Action Selection for Robots in Dynamic Environments through Interbehaviour Bidding, in Dave Cliff et al., eds., Proceedings of the Third International Conference on Simulation of Adaptive Behavior (SAB-94).

....of agents for such maintenance tasks, SUMPY promises to prove useful, and has added no significant problems to the test systems. Key words: autonomous agent, fuzzy controller, software agent, subsumption architecture. 1. Introduction Autonomous agents have shown considerable promise in robotics [1,2,3,12,13], software development [5] information mining in network environments [4,10] life science [11] and in other fields. Thus control structures for such autonomous agents raise important research issues. What are the useful architectures for what tasks Many interesting control structures have been ....

Sahota, Michael K. (1994), Action Selection for Robots in Dynamic Environments through Inter-behavior Bidding, in Dave Cliff et al., eds., Proceedings of the Third International Conference on Simulation of Adaptive Behavior (SAB-94).

....whatever actions the agent has learnt to take to pursue its goals. It is where the W values W i (x) come from, and how they change in response to not being obeyed, that within a single body, Brooks, 1986] uses layer (though [Brooks, 1994] also uses process) Blumberg, 1994] uses activity and [Sahota, 1994] uses behavior. is the interesting bit. Schemes using such importance values are common in multi behavior models (e.g. see the utility functions in [Aylett, 1995] and are normally hand designed. To get them generated for free, we look to reinforcement learning. 2 Reinforcement Learning ....

....we wanted was a strict hierarchy, there would be no problem to solve. It is clear that in many situations we do not want a strict hierarchy but rather want to see opportunistic behavior, where different goals are partially satisfied on the way to solving other goals. Dithering [Minsky, 1986, Sahota, 1994] in general is avoided in W learning collections since agents can tell the difference between situations when they are likely to get an immediate payoff and situations when they could only begin some sequence of actions which will lead to a payoff later. The agents will present different W values ....

Sahota, Michael K. (1994), Action Selection for Robots in Dynamic Environments through Inter-behaviour Bidding, in Dave Cliff et al., eds., Proceedings of the Third International Conference on Simulation of Adaptive Behavior (SAB-94).

....The inter relationship of these behaviors can perform tasks such as hall following etc. However in some cases several contradictory behaviors may be invoked simultaneously and arbitration is required. The arbitration can be a fixed hierarchy as in (Brooks 1986) or by inter behavioral bidding as in (Sahota 1994) or by a variety of other schemes. A fixed hierarchy can be quite brittle especially when the robot is tested in an environment different to the one in which it was developed. Of particular relevance to our work are techniques for shared control between robot and user. This research is common in ....

Sahota M.K. 1994. Action selection for robots in dynamic environments through inter- behaviour bidding. In Proceedings Simulation of Adaptive Behaviour, 138-142, Brighton, England.

....giving them full sensingand acting powers, but does not go as far as letting them compete for control. Instead, action selection is a job for the programmer. Brooks original scheme has been extended (see survey in [Brooks, 1994] and other schemes have been proposed [Maes, 1989, Blumberg, 1994, Sahota, 1994] but action selection remains something basically designed rather than self organised. In an attempt to avoid this problem of design, I introduce a model in which yet further liberation of modules is attempted. 1.2 Competition among selfish agents I introduce a model (Figure 2) with the ....

....inside it. For somewhatautonomous, somewhat competing modules within a single physical robot, Brooks, 1986] uses layer (though [Brooks, 1994] also uses process) Minsky, 1986] uses agent (even though most of his agents do not interact directly with the world) Blumberg, 1994] uses activity and [Sahota, 1994] uses behavior. To be more precise, let the collection consist of agents A 1 ; An . Time steps are discrete. Each time step, the robot observes the world to be in some state x. Each agent A i suggests an action a i (x) that it wants to see executed in this state. The robot chooses ....

Sahota, Michael K. (1994), Action Selection for Robots in Dynamic Environments through Inter-behaviour Bidding, in Dave Cliff et al., eds., Proceedings of the Third International Conference on Simulation of Adaptive Behavior (SAB94) .

....in dynamic environments [21] The relationship of these behaviours can perform tasks such as hall following etc. However in cases several behaviours my be invoked simultaneously and arbitration is required. The arbitration can be a fixed hierarchy as in [21] or by inter behavioural bidding as in [14] or by a variety of other schemes. A fixed hierarchy can be quite brittle especially when the robot is tested in an environment different to the one in which it was developed. Of particular relevance to our work are techniques for shared control between robot and user. This research is common in ....

Sahota Michaek K. Action selection for robots in dynamic environments through interbehaviour bidding. In Proceedings SAB-94, Brighton, August 1994.

No context found.

Sahota, Michael K. (1994), Action Selection for Robots in Dynamic Environments through Inter-behaviour Bidding, in Dave Cliff et al., eds., Proceedings of the Third International Conference on Simulation of Adaptive Behavior (SAB94) .

No context found.

M. Sahota. Action selection for robots in dynamic environments through interbehaviour bidding. In [SAB94].

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