Complete
Coverage of Unstructured Environments
Path-planning algorithms are well understood for a variety of exploratory applications. Artificial-intelligence methods like learning algorithms are widely used to solve the problem of robot navigation in unstructured environments. However there are some types of mobile-robot applications that need a different path-planning technique. Scanning applications, like landmine detection, cleaning tasks and terrain-map generation, require not finding the shortest path to a point in the environment but scanning over all points in the environment and avoiding obstacles of unknown location. This is known as the complete-coverage problem in unstructured environments. Our goal is to achieve complete coverage of unstructured environments for landmine detection purposes. DYLEMA project (see Figure 1) is about considering walking robots as locomotion system that carry a scanning system to detect and locate buried mines. Therefore we need to achieve complete coverage in the navigation of both walking robot and scanning system.
|
Figure 1: DYLEMA: a walking
platform for antipersonnel mine detection |
1. Mobile robot navigation with complete coverage of unstructured environments: A variety of coverage algorithms exist, but the most powerful ones are those that rely on finding the critical points of a function to guarantee the completeness of the method. The complete-coverage algorithm herein proposed is of this type and extends Choset's algorithm to achieve better performance in unstructured environments.
Garcia, E. and Gonzalez de Santos, P. Mobile Robot Navigation with Complete Coverage of Unstructured Environments. Robotics and Autonomous Systems, vol.46, no. 4, pp. 195-204, 2004.
2. Hybrid Deliberative/Reactive Control of a Scanning System for Landmine Detection: These are the on going results about the scanning manipulator carried by the walking robot. The scanning manipulator is a 5-DOF robot manipulator endowed with a sensor head (see Figure 2(a)). The sensor head contains a commercial metal detector to locate buried mines, and range sensors used to control the detector motion to follow the ground contour (see Figure 2(b)). A simple array of bumpers (flex sensors) around the sensor head detects obstacles in the scanning trajectory. The on-board controller modifies on-line the search trajectory to avoid obstacles like rocks or trees. Therefore, the scanning system allows to autonomously search for landmines in unstructured, natural terrains. The control architecture herein presented is of the deliberative/reactive hybrid type, thus commanding a predefined sweep trajectory that ensures the complete coverage of the swept area, but modifying it in a reactive manner to adapt to terrain roughness and avoiding obstacles in the path. The complete-coverage sweep motion of the scanning system is original of this work. The use of the proposed reactive controllers allow an efficient operation in natural, unstructured environments without resorting to complex and expensive methods for on-line terrain mapping.
|
|
Here are some videos of the scanning system:
Video1: Sweep trajectory (12 MB-mpeg file)
Video 2: Ground-contour tracer (14 MB-mpeg file)
Video 3: Object-contour tracer (6MB-mpeg file)