Behaviors for Non-Holonomic Box-Pushing Robots

We describe a behavior-based control system that enables a non-holonomic robot to push an object from an arbitrary starting position to a goal location through an obstacle field. The approach avoids the need for maintaining an internal model of the target and obstacles and the potentially high computational overhead associated with traditional path planning approaches for non-holonomic robots. The motor schema-based control system was prototyped in simulation, then tested on mobile robots. The approach was demonstrated in two pushing tasks: box pushing and ball dribbling. Using the same generalized control system, a robot is able to successfully push a box through a static obstacle field and dribble a ball into a goal during RoboCup soccer matches. In both cases the robot is able to react quickly and recover from situations where it loses control of its target either due to its own motion or interference with others. Quantitative xperimental results for reliability in box pushing are included. Background and Related Work Pushing is the basis for many useful tasks. In a construction scenario, for instance, bulldozer robots are essentially performing a pushing task in which the target object is soil (Singh & Cannon 1998). Pushing is also important in other materials handling tasks such as assembly, construction and cleaning. For pushing, a robot does not require lifting or gripping mechanisms and therefore is more tolerant to load size and shape. Also, heavy, unliftable, objects can be transported by multiple robots coordinating their pushing efforts (Mataric, Nilsson, & Simsarian 1995; Parker 1994). Several of the "sub-skills" necessary for pushing are also useful in other tasks. For example, pushing requires a robot to align itself properly at the target with respect to a goal location and then maintain control of that object while travelling to the goal. The alignment phase of pushing can be used to properly position a robot for picking up a target. If the robot is non-holonomic, this alignment phase is non-trivial. A number of researchers have investigated path and push planning for non-holonomic robots; however, these solutions tend to address static environments and they are computationally more expensive than our approach (Latombe Copyright © 2000, American Association for Artificial Intelligence (www.aaai.org). All rights reserved. 1991; Lynch & Mason 1996). Additionally, traditional path planning approaches for a pushing task may require accurate models of the environment or other global information that is not easily available to a robot in real applications. One alternative to path-planning is a behavior-based approach. This family of methods closely tie perception to action which makes them appropriate for dynamic environments or environments for which no a priori world model exists. In this work we use a motor schema-based approach which combines behavioral building blocks using vector summation similar to potential field navigation methods (Arkin 1989; Latombe 1991). Ours is not the first behavior-based approach to box pushing. Parker, and separately, Mataric, for instance, have investigated coordinated multi-robot box pushing (Mataric, Nilsson, & Simsarian 1995; Parker 1994). However, ours is the first to address the problem of pushing while avoiding obstacles and simultaneously accounting for the nonholonomic onstraints of a tricycle-like robot.