3 rd International Symposium on Electrical Engineering and Energy Converters September 24-25, 2009, Suceava 115 Abstract— Control of legged robots can be inspired from the way in which biological systems (living creatures) control the movements. This paper deals with the problem of shape memory alloy spring based legged robot control in measurable but unpredictable environments. The paper structure consists of two sections: first section studies the use of shape memory alloy leg structure and the second section deals with the evolution performed using a causality structure with four free joints, with desired values for the centre position and for the body angle of the robot, codified as causality structure [motor 15 motor 25 4]. All the researches developed until now, for the robot represented as a variable causality dynamical system (VCDS), are kept and used for the causality structure approached in this paper. The results are implemented and verified in RoPa, a platform for simulation and design of walking robot control algorithms and some evolution examples are presented. Index Terms—causality structure, control algorithms, desired trajectories, shape memory alloy, legged robot I. INTRODUCTION Recently, intensive studies have been focused on legged robots. Compared to traditional wheeled robots, walking robots will be able to handle uneven terrain and soft ground in difficult conditions where wheeled robots cannot go. Furthermore, one can take the advantages of biologically inspired control strategies and apply the control scheme to robots through observing how living creatures control their movements. Behavior of walking robots from the biped structure till the multileg structures is characterized by a specific type of movement called legged locomotion, [15], [6]. First of all, the robot leg has to offer not only a sure contact surface, but an adaptive damper coefficient in order to adapt the robot movement to unknown environment. A simplified model for springy robot leg is assimilated with a pogo stick – Fig. 1. The variables in the model are positions and velocities, and the dynamic equations come from Newton’s laws of motion. When humans walk, feet never lose contact with the ground and alternate between having both feet on the ground and a swing phase in which one foot is on the ground and the second leg swings like a pendulum. When run, we alternate between a flight phase in which both feet are off the ground and a stance phase in which one foot is on the ground. Kangaroos hop with a flight phase alternating with a stance phase in which both feet are on the ground simultaneously. Fig. 1 A simple model for running and hopping leg As an approximation, one can think of the springy leg as the tendons in the leg. By contracting muscles, the animal changes the force of the leg spring, enabling it to bounce off the ground. When in flight, we assume that the animal is able to swing the leg so that it will point in a new direction when the animal lands on the ground. At landing, the leg shortens, compressing the spring. The compressed spring exerts a vertical upward force that together with additional force exerted by the muscles propels the animal into its next flight phase. Simplifying, the hopping movement can be divided in hopping in place and hopping forward or backward. Hopping in place and hopping forward/backward can both be divided into two different phases: an aerial phase (where the mass is airborne) and a ground phase (where the mass and spring are on the ground). Control algorithms for legged locomotion are very different but all of them must assure a stable movement. From this point of view, there are two types of stable movements: dynamic stable movement and static stable movement. Many control algorithms implemented on the existing walking robots, [7], are based on "state of the art" technologies to control the movements of articulated limbs and joint actuators. Some of them try to recreate the efficient yet very complex movements of biological insects and mammals, which effortlessly execute various types of periodic gait patterns and adaptive gaits at very high speed [6]. Usually, two-legged robots are designed according to the human skeleton and controlled according to human behaviors. This encourages many researchers to investigate the basic human movements and try to apply the human behavior to robots. Today, the notion of central pattern generator (CPG) has been developed based on living organisms to generate the human-like gait rhythm. The newest research of biologically inspired walking machine Biomimetic Approach in Developing Control Strategy for Shape Memory Alloy Legged Robot Anca PETRISOR, Nicu George BIZDOACA, Mircea Adrian DRIGHICIU, Raducu PETRISOR University of Craiova Blv. Decebal nr.107, RO-200440 Craiova apetrisor@em.ucv.ro, nicu@robotics.ucv.ro,adrighiciu@em.ucv.ro, raducu.petrisor@elprest.ro