5. ADVANCED DESIGN AND CONTROL
5. ADVANCED DESIGN AND CONTROL
5.1. The INRIA Project SAGA
Activities of the INRIA SAGA project in the field of robotics are related to the development of new robot mechanisms, especially parallel manipulators, optimal design and the use of equation solvers and symbolic computation for motion planning.
The mechanical architecture of parallel robots offers some interesting performances like accuracy, rigidity, high nominal load and may constitute an alternative to serial robots, for example for fine positioning devices or medical robotics. But this type of mechanisms have to be carefully studied as the involved theoretical problems are quite different from the classical ones involved for serial robots. Hence various topics are investigated like workspace and singularity analysis, direct kinematics, motion verification, force-feedback control. To deal with these topics the approach is first to check carefully the geometry of the problem in order to get the best insight on how to solve it. For example in workspace analysis it has been possible to show that the workspace border may be obtained, in general, as the intersection of geometrical objects that can easily be computed: this leads to very efficient, almost real-time, algorithms.
These results have been validated on two prototypes and lead to the development of a new parallel micro-robot (with a diameter of 7 mm) which is intended to be used as an active head for medical and industrial endoscopy.
For complex mechanisms as, for example, parallel robots the dimensions defining the geometry of the robot have a large influence on the performances. Hence the interest to look for a methodology of optimal design i.e. aiming to find the dimensions of the robot such that the performances of the machine will fit in the best way the requirements of the end-user. Classical optimal design methods like the cost-function approach cannot, in general, be used for such problem since the number of parameters and requirements is too large. The proposed approach is first to develop tools for performance analysis. These tools enable the designer to determine efficiently and with guaranteed results what will be the performances of a given robot (e.g. find the worst positioning error in the whole workspace of the robot). This is usually a complex problem as these performances are in general varying with the pose of the robot, while the designer can be mainly interested in the worst or best cases over the whole workspace.
The method in the second step tries to reduce the possible values of the design parameters. This task relies on the concept of the parameter space, a n-dimensional space in which each dimension represents one of the design parameters (hence a point in this space represents a unique robot geometry). Next step is to consider a requirement and try to determine the region of the parameter space such that all the robots that fulfil the requirement will have a representative point in this space that is included in the region. For example it has been developed an algorithm that determines such region when the requirement is that a set of poses has to belong to the workspace of the robot. The regions are calculated for various types of requirements and the optimal design has to belong to the intersection of all these regions.
As soon as this intersection has been computed, it is discretized, and for each node of the grid, the performance analysis tools are used to verify if the robot satisfies at best all the design requirements, thus allowing to determine a quasi-optimal design.
Another of the objectives of the SAGA project is to develop efficient methods for analysing and solving systems of equations, a problem that occurs quite frequently in robotics, for example in kinematics or motion planning. These methods have been used for the optimal design problem and for improving the motion planning algorithm proposed by Canny.
5.2. The BIP 2000 Anthropomorphic Biped Robot
BIP 2000 is an INRIA project coordinated by the BIP team of INRIA Rhône-Alpes and carried out jointly with three other laboratories: LAG, Grenoble ; LMS and LMP, Poitiers.
This project is aimed at the realization of the lower part of an anthropomorphic biped robot. The interest in these robots lies in their natural capability for operating in the essentially bipedal-friendliness of our everyday environment. The priority class of applications will be that of service robots, with the hope of further spin-offs in the area of biomechanics.
The project covers mechanical design, control studies and computer architecture integration.
The links geometry, the mass distribution, the kinematics and the capacities of the robot in terms of joint torques and velocities are close to the ones of humans.
The transmissions are specific screw/nuts-based systems which have good dynamic performances, reversibility and small size. They are arranged in parallel at the ankles and at the trunk/pelvis linkage.
The control schemes are either static, based on a control of the center of mass associated with suitable task functions, or dynamic, taking into account the unilateral constraints foot/ground.
In the present state of the project, the salient results obtained are:
The first version of the BIP 2000 robot, called BIP1, has two legs with 8 degrees of freedom. BIP1 took its first steps in March 2000.
The second version, called BIP2, consists of two legs, a pelvis and a trunk with 15 degrees of freedom. BIP2 can maintain its static balance on one foot while moving the other leg.
The software environments for real-time control of the system is complete and operational. It is called ORCCAD.
The robot is being presented in Hannover at the Expo2000 to illustrate a prototype biped robot capable of exhibiting human-like locomotion.