Bilkent University
Department of Computer Engineering


State Estimation on SLIP using an Approximate Motion Model


Özlem Gür
Ph.D. Student
Computer Engineering Department
Bilkent University

Robotics, being a rapidly evolving area, helps human kind not only by handling everyday tasks in an automated manner but also attempting to perform tasks which cannot be done manually. Today, robots can be used in various applications such as household work, military applications, space exploration, assembly lines in factories and automated guided vehicles. Depending on the application, different kinds of robots should be selected. Legged robots are better suited for outdoor applications due to their ability to operate in uneven terrain at high speeds. Moreover, application areas of wheeled vehicles in the world are limited to structured arenas (e.g. roads and rails) and some natural ambulatories. The simplest and most fundamental leg model for humans, animals and robots is Spring Loaded Inverted Pendulum (SLIP) which consists of a body mass and a massless spring for the leg. Among numerous research areas based on SLIP model, we focus on state estimation which is used for localization, mapping and sensor fusion. State estimation deals with the robot's perception of its own state and the environment in the existence of noisy measurements and control inputs. Without having a sense of its own state (position, velocity, etc.), many tasks such as traveling to a goal configuration, interacting with the environment, applying control and vision algorithms on the robot are not feasible. In this study, the application of Extended Kalman Filter, a state estimation technique, on SLIP model will be presented. Moreover, the effects of using an approximate motion model on the performance of the filter will be investigated and the results will be discussed based on simulation results.


DATE: 26 April, 2010, Monday @ 15:40