Abstract | ||
---|---|---|
In this paper, we explore the idea of using inertial and actuator information to accurately identify the environment of an amphibious robot. In particular, in our work with a legged robot we use internal sensors to measure the dynamics and interaction forces experienced by the robot. From these measurements we use simple machine learning methods to prob- abilistically infer properties of the environment, and therefore identify it. The robot's gait can then be automatically selected in response to environmental changes. Experimental results show that for several environments (sand, water, snow, ice, etc.), the identification process is over 90 per cent accurate. The requisite data can be collected during a half-leg rotation (about 250 ms), making it one of the fastest and most economical environment identifiers for a dynamic robot. For the littoral setting, a gait- change experiment is done as a proof-of-concept of a robot automatically adapting its gait to suit the environment. I. INTRODUCTION In this paper we demonstrate adaptive gait control to a wide range of environments for a legged robot. In particular, we demonstrate that inertial sensors and actuator feedback are sufficient to leverage a Bayesian classifier that rapidly identifies the environment, despite large amounts of noise and intermittent contact. This information then allows the robot to chose its gait both qualitatively and quantitatively to adapt to the current environment. Furthermore, we believe this is the first work that demonstrates the efficiency of such methods over such a wide range of environmental contexts including swimming underwater, walking on slippery ice, and traversing the open spaces of a typical university office complex. Practical implications of this include, for example, the ability of the robot to switch from walking to swimming gaits as it moves from a sand beach or surf-zone to deep water. Our experimental testbed, AQUA, is an amphibious hexapod with six independently-controlled leg actuators. The robot can negotiate rugged terrains, and with the use of amphibious legs, it can also swim in water to a depth of 10 m. Proper selection of gait for each type of environment is of crucial importance. Therefore any autonomous version of the robot would have to identify the environment in order to select the proper gait. Since for this robot the leg forces are by nature very impulse-like, the robot dynamics highly depend of the surface mechanical properties. Conveniently, this behavior can be viewed as a mechanism for probing and estimating the dynamic properties of the surface. Fig. 1. The hexapod robot, shown equipped with the semi-circle legs. Communication with the operator station occurs over the fiber-optic tether. |
Year | Venue | Keywords |
---|---|---|
2006 | Robotics: Science and Systems | environmental change,fiber optic,surf zone,machine learning,proof of concept,bayesian classifier |
Field | DocType | Citations |
Inertial frame of reference,Computer vision,Robot calibration,Simple machine,Gait,Identifier,Simulation,Computer science,Legged robot,Artificial intelligence,Robot,Actuator | Conference | 16 |
PageRank | References | Authors |
0.81 | 12 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Philippe Giguère | 1 | 145 | 21.51 |
Gregory Dudek | 2 | 2163 | 255.48 |
Shane Saunderson | 3 | 68 | 5.93 |
Chris Prahacs | 4 | 68 | 6.26 |