3. EXTREME ENVIRONMENTS AND FIELD ROBOTICS
3. EXTREME ENVIRONMENTS AND FIELD ROBOTICS
3.1. Remote Hostile-Sites Intervention
Field Robotics at LAAS-CNRS
The research effort on long-range navigation is continuing, and the cooperation with two industries, Aerospatiale and Alcatel Space is coming to its term. The mobile robot LAMA, equipped with two stereo-pairs of cameras, a gyrometer and inclinometers is used for the experimental developments in the EDEN site et LAAS. This work concerns mainly:
"optical odometry", i.e., robot localization using setreovision and pixel tracking.
Reactive navigation on uneven terrain with visual terrain mapping into an elevation map and motion generation.
Locomotion control using a model of wheel-soil friction for traction control.
Landmark tracking for visually guided navigation.
At a longer term, it is considered to investigate the use of a blimp exploring the terrain in coordination with the terrestrial robot.
The data acquired during the November-December 1998 expedition to Patriot Hills, Antarctica, in cooperation with Carnegie Mellon University, where processed for understanding the limits of image processing algorithms and laser data exploitation in extreme environments. This resulted in investigating new approaches for vision in such conditions.
With Carnegie Mellon University, USA, exchange of students was done.
A cooperation program with the Australian Center for Field Robotics of the University of Sydney, Australia, has been launched. It includes exchange of students and software for field robotics.
3.2. Smart Vehicles and Multi-Sensor Systems for Humanitarian Demining
The Laboratoire de Robotique de Paris (LRP) and the Ecole Nationale Supérieure d’Ingénieurs de Bourges (ENSIB) develop robotics research projects in the field of smart detection technologies addressing the needs of humanitarian demining. Particular focuses of these projects are on the sensor technologies, their mobility for the ground exploration and learning techniques for detection.
Integrated multi-sensor solutions, strengthened by data fusion are part of the work aims. This research is undertaken by ENSIB in a close relationship with industrial partners in the "Pôle Capteur de la Région Centre" framework. Current investigations are on a scenario-configurable smart multi-sensor head integrating metal detectors, electromagnetic sensors, ground penetrating radar, micro-wave sensors, an X-ray sensor and an "artificial nose". The hardware integration of these sensors and the data/signal processing are performed to improve significantly the speed and the efficiency of the detection.
The multi-sensor head will be integrated in an innovative off-road vehicle. A research action developed at LRP concerns the mechanical design of rovers with extended terrain capabilities to access different terrain topologies and terramechanic properties. To achieve this high "terrainability", a powerful design tool based on Adaptive Evolutionary Algorithms intregrated in a global evaluation of the system dynamic behavior has been developed. Different designs of locomotion mechanisms combining legs and wheels have been created. An original hybrid mechanical architecture which consists in four wheels, independently driven and streerable, connected to the platform by two degrees of freedom mechanisms has been prototyped. Inclinometers and force sensors provide informations about the geometrical and mechanical interactions of the vehicle with the ground. The high mobility and the actuation redundancy provide to the system the capacity to increase the stability and the control of the platform configuration, to optimize load and traction force balance, to use a peristaltic mode The adaptivity of the locomotion mode relatively to the terrain characteristics obtained from observation of the state of the vehicle has been investigated in cooperation with the Laboratoire d'Automatique et d'Analyse des Systèmes (LAAS) and Russian teams involved in the Centre Associatif de Recherche en Robotique Avancée (CARRA) framework. A particular attention has been paid to the control of interactions between the system of wheels and natural soils. Different analytical models (including elasto-plastic soils and sand) have been proposed and evaluated experimentally on the LAMA vehicle for the control of wheel slippage to increase the traction and the odometry performances.
Future works in this research project will consider the multi-sensor head integration in the vehicle and the development of fast detection strategies of APLs while preserving the vehicle integrity.
3.3. Construction and Civil Work Robotics
The activity in the field of Construction and Civil Work Robotics in France has been federated in the last years by IRCCyN, Institut de Recherche en Communication et Cybernétique de Nantes, in the framework of the CNRS Strategic Action on Intelligent Machines, on the basis of existing cooperative actions. Basically, this pole has been concerned by the partial automation of road construction machines, such as the paver and the compactor.
The first large scale action in this activity was the EU Brite-Euram project CIRC (Computer Integrated Road Construction), piloted by ITMI, which came to an end in January 2000. The main idea of CIRC is to fill the gap in Information Technology between the highly computerized design phase of the road and the building phase, where IT is hardly represented.
Two operations have been chosen to exemplify the high potential of improvement through the use of IT:
The operation of paving (applying the final layer on the road). Here the idea is to use the CAD definition of the desired road surface to control the tool of the paver (a vibrating table, of which the altitude of the left and right ends can be controlled) in order to apply the correct amount of material on the previous layer.
The operation of compaction. This operation is performed by compactors driven by an operator. The operator has to perform a number of passes on each point of the road to achieve the correct density of the material, in accordance with a predefined plan. In this case, the idea is to keep track of the passes performed at each point and of the current position of the compactor to help him in his task. Recording the passes performed also helps as a quality control tool.
In both of these tasks, the localization function has been identified as the keystone of automation and has been one of the main concerns of the CIRC project. One of the deliverables of the project was the CIRPAV (CIRC product for pavers) system. CIRPAV is made up of two essential components : a localization system (an automatic theodolite called Laserguide) capable of providing the six degrees-of-freedom of a low-speed vehicle in real-time, and a tool control algorithm to track the time-variable desired position and attitude necessary to generate the correct road profile. A second product was the CIRCOM (CIRC for compactors) product, which entails both a localization system for the compactor and a graphical user-interface to help the driver keep track of his work.
Although the CIRC project has been a real step forward in these fields, a few shortcomings of the developed solutions also have appeared. The requirements for CIRPAV are a precision of 1 cm in altitude, 5 cm in x and y, 0.1° in attitude with respect to the horizontal plane and 0.5° in heading. These performances are not fully satisfied by the current version of the localization system. This is the reason why a cooperative work is being performed between the University of East London (UEL), designer of the hardware and first localization software), Laboratoire Central des Ponts et Chaussées (LCPC), which has developed a special-purpose hardware for testing localization systems and IRCCyN, designer of new localization algorithms. New tests performed at LCPC show that, with the new software, the requirements defined for the CIRC project are achieved.
Some of the actors of this work, together with new industrial partners, are currently preparing a proposal for an Exploratory Award, as a framework for the development of a final industrial version of the Laserguide system.
Simultaneously to the development of the paver localization system, cooperative work between LCPC and IRCCyN has been done to improve the GPS-based localization of compactors defined with the framework of CIRC. IRCCyN, who developed the localization algorithms of CIRCOM as a sub-contractor, has developed algorithms based on less expensive sensor configurations, while retaining the capability of the system to operate during moderately long masking phases of the GPS, due to working under a bridge for example.
LCPC and IRCCyN also have developed longer-term cooperation around the modeling and identification of the dynamics of compactors. The aim of this work is to allow later the automatic control of the compactor, with the operator having essentially a security role. Another goal is, via the knowledge of the dynamic model which takes into account the interaction between the cylinders of the compactor and the ground, to estimate the density of the material. This would allow to define the task directly as a desired density, rather than transforming it into a certain number of passes using more or less empirical laws.
This last goal is one of the objectives of the recently started Brite-Euram European project OSYRIS (Open System for Road Information Support). The goal of OSYRIS is to develop and validate a line of products based on a common Information Technology infrastructure, which will allow to trace the road life cycle by measuring and collecting data, processing them, and making them available in real-time to critical actors on-site and off-site. The online estimation of density by the compactor in OSYRIS aims at illustrating the use of novel, intelligent sensors and the data collection aspects.
In order to implement online identification of the density, research on the modeling of the compactor currently focuses on taking into account in the dynamic model the vibrations applied to the cylinders of the machine to efficiently compact the material. Until now, models without vibration had been developed and partially identified. Identification of the model is particularly difficult because the skidding and slipping of the cylinders with respect to the ground are required, and they are difficult to estimate.