RPS is a path generator for autonomous nonholonomic mobile robots in partially known aisle environments. Autonomous vehicles are vehicles that are capable of intelligent motion and action without requiring either a guide to follow or teleoperator control. RPS is designed for autonomous vehicle obstacle avoidance and utilizes information about its environment through ultrasonic range sensors.
RPS derives its name from its ability to "react" to local surroundings. A path is constructed one piece at a time, based upon local environmental information. Each piece is initially considered to be rough and needs to be appropriately "shaped," or smoothed, to meet mobile robot nonholonomic kinematic constraints.
A nonholonomic mechanism has fewer local degrees of freedom than global. Because of this limitation, or constraint, a path generated for a holonomic vehicle is often impossible for a nonholonomic vehicle to follow.
A piecewise approach to path generation is taken due to the kinematic constraints of the nonholonomic mobile robot. Difficulties arise when using a perfectly reactive path generator, such as accounting for steering angle constraints. For example, a perfectly reactive path generator may call for a turn that is too "sharp," and beyond the robot's turning capabilities. An autonomous vehicle path must also meet curvature continuity constraints, to insure that it is feasible for the nonholonomic mobile robot.
RPS uses a method for collision avoidance that is similar to many potential field methods. In general, potential field methods apply repulsive forces from obstacles to the vehicle itself which may result in paths that require drastic configuration changes for nonholonomic vehicles. RPS applies repulsive forces to a part of the path ahead of the vehicle so there is time to make the path suitable for a nonholonomic vehicle. This thesis assumes static obstacles.
RPS constructs curves from straight line segments and segments of a special class of curves called clothoids. Clothoids are continuous curves whose curvature varies linearly along its arc length. Clothoids have been used by researchers to satisfy constraints of nonholonomic mobile robots.2 RPS utilizes an environmental map to keep track of its surroundings. The environmental map efficiently stores and manipulates the large amount of range data from ultrasonic sensors. It is from this map that RPS generates a kind of potential field, keeping the path away from sensed obstacles.
RPS was developed utilizing the Multi-Body Simulator (MBSim), an experimental simulation platform that was created at Georgia Tech. MBSim has the ability to model, simulate, and animate the kinematics and dynamics of different types of mobile robots with range sensors. Several test environments for RPS were developed in MBSim and data is presented on the results of these experiments.
The chief application that drove the development of RPS is autonomous warehouse inspection. Of particular interest is the inspection of barrels containing hazardous compounds and lined in aisles of pallets. To make the most of warehouse space, the aisles tend to be narrow. In addition, the placement and regularity of aisles tend to vary as barrels are added and removed over time. A primary concern is avoiding collisions between the mobile robot and barrels, deterring expensive repairs or clean-up operations. These factors present interesting challenges to autonomous robots. Usually a list of goal points, namely aisle intersections, is known, but, due to aisle irregularities, the path between the points may not be a straight line. It is to this endeavor that RPS has been applied.
RPS has significant difficulty with long obstacles perpendicularly placed between a vehicle and its goal point. Such an obstacle affects the path little, and as a result, the vehicle attempts to drive through the obstacle. This is a significant limitation to RPS and prevents it from being used as a general path planner.
If you have a Silicon Graphics machine, download RPSDemo.tar.gz which uses GL graphics. (126k)
If you are interested in learning more, PLEASE EMAIL ME!!! Also, you are free to download a copy of my thesis. It is a 147 page doument broken down into seven chapters. The above text was Chapter 1!
Chapter 2 summarizes other researchers' work related to RPS. Chapter 3 gives a brief description of nonholonomic vehicles. It also looks at requirements for nonholonomic vehicle paths, and why clothoids are particularly useful. Chapter 4 covers RPS in detail how the rough path is generated and then smoothed. Chapter 5 describes the simulation platform used in the development of RPS, MBSim. A familiarization of MBSim is necessary to understand Chapter 6, a description of experiments and their results used to test RPS. Chapter 7 summarizes this work and suggests a plan for future work.
The thesis is broken into 3 encapsulated postscript files. Warning: Together these files sum to approximately 4 megs! (Note: Email me for the Word 6.0 documents as I am having difficulty posting them directly here.)
Last Updated: February 8, 1996