Title :
Learning tactic-based motion models with fast particle smoothing
Author :
Gu, Yang ; Veloso, Manuela
Author_Institution :
Sch. of Comput. Sci., Carnegie Mellon Univ., Pittsburgh, PA
Abstract :
Learning parameters of a motion model is an important challenge for autonomous robots. We address the particular instance of parameter learning when tracking motions with a switching state-space model. We present a general algorithm for dealing simultaneously with both unknown fixed model parameters and state variables. Using an Expectation-Maximization approach, we apply a tactic-based multi-model particle filter to estimate the state variables in the E-step, and use particle smoothing to update the parameters in the M-step. We test our algorithm both in simulation and in a team robot soccer environment, as a substrate for applying the learned models to object tracking in a team. One of the soccer robots learns the actuation model of its teammate. The experimental results show that the particle smoothing efficiency is substantially increased and the tracking performance is significantly improved using the learned teammate actuation model.
Keywords :
expectation-maximisation algorithm; learning (artificial intelligence); mobile robots; multi-robot systems; object detection; particle filtering (numerical methods); robot vision; smoothing methods; sport; state-space methods; autonomous robot; expectation-maximization approach; learned teammate actuation model; object tracking; particle smoothing; switching state-space model; tactic-based motion model parameter learning; tactic-based multimodel particle filter; team robot soccer environment; Bayesian methods; Computer science; Humans; Particle filters; Robot kinematics; Robot sensing systems; Robotics and automation; Smoothing methods; State estimation; Tracking;
Conference_Titel :
Robotics and Automation, 2008. ICRA 2008. IEEE International Conference on
Conference_Location :
Pasadena, CA
Print_ISBN :
978-1-4244-1646-2
Electronic_ISBN :
1050-4729
DOI :
10.1109/ROBOT.2008.4543720