A Novel Distributed Prediction Market Model and Algorithm for Forecasting Outcomes of Related Events
Advisor Information
Prithviraj Dasgupta
Location
Milo Bail Student Center Omaha Room
Presentation Type
Oral Presentation
Start Date
8-3-2013 11:45 AM
End Date
8-3-2013 12:00 PM
Abstract
In this research we consider the problem of automatically recon_guring or changing the shape of a modular self-recon_gurable robot (MSR) when it cannot continue its motion or task in its current shape. To solve the modular robot recon_guration problem, we propose a novel technique based on a branch of economics called coalition game theory, which is used by people to divide themselves into teams or coalitions. The conventional computer algorithm used for forming coalitions and _nding the best coalitions is very expensive to implement in terms of running time and energy (battery power) and not practical to implement on small-scale, modular robots. We have proposed a new, fast algorithm called searchUCSG that intelligently reduces the number of coalitions it needs to inspect and eventually _nds the best coalitions for the modules of the modular robot. Our proposed technique also incorporates an essential aspect of robotics- uncertainly in operation of the robots movements. We have veri_ed the operation of our algorithm mathematically as well as experimentally using a computer simulated model of a modular robot called ModRED that we are developing as part of the NASA-sponsored ModRED project. Experimental results of our algorithm show that it is able to recon_gure a modular robot while taking signi_cantly lesser time than other state-of-the-art algorithms and is able to form a con_guration that is very close or at worst 80% away from the best possible con_guration of the modules.
A Novel Distributed Prediction Market Model and Algorithm for Forecasting Outcomes of Related Events
Milo Bail Student Center Omaha Room
In this research we consider the problem of automatically recon_guring or changing the shape of a modular self-recon_gurable robot (MSR) when it cannot continue its motion or task in its current shape. To solve the modular robot recon_guration problem, we propose a novel technique based on a branch of economics called coalition game theory, which is used by people to divide themselves into teams or coalitions. The conventional computer algorithm used for forming coalitions and _nding the best coalitions is very expensive to implement in terms of running time and energy (battery power) and not practical to implement on small-scale, modular robots. We have proposed a new, fast algorithm called searchUCSG that intelligently reduces the number of coalitions it needs to inspect and eventually _nds the best coalitions for the modules of the modular robot. Our proposed technique also incorporates an essential aspect of robotics- uncertainly in operation of the robots movements. We have veri_ed the operation of our algorithm mathematically as well as experimentally using a computer simulated model of a modular robot called ModRED that we are developing as part of the NASA-sponsored ModRED project. Experimental results of our algorithm show that it is able to recon_gure a modular robot while taking signi_cantly lesser time than other state-of-the-art algorithms and is able to form a con_guration that is very close or at worst 80% away from the best possible con_guration of the modules.