Vladimir Kishinskiy

Faculty:

Speciality:

Software of the system

Theme of master's work:

The development of intelligent mobile robot motion control

Scientific supervisor:

Candidate of technical sciences, associate professor of department SIS

Elena Volchenko

ABSTRACT

• Introduction
• 1. Purposes and tasks
• 2. Analysis of existing methods
• 3. Conclusions
• The list of the used sources

Many research teams are engaged in research of adaptive walking machines and management of their movement. In comparison with traditional wheel and track laying vehicles walking machines have the best characteristics on passableness when walking on weak soil and when moving on a surface with a difficult relief. The modern technological level and computer facilities allows to build independent machines. They can serve for performance of concrete technological operations on the difficult district [10].

Now level of automation and intellectual systems allows to build independent systems. Use of the operator is sometimes inexpedient or it is impossible, therefore the developed system should be completely automatic.

Object of research is the walking robot or the walking chassis for equipment.
Subject of research is the algorithm of formation of gait and movement of extremities of the robot.

Moving of the walking robot is characterized:

• Complexity of the exact description of model of movement because It have a large number of influencing factors;
• A large number of parameters which influence process of walking;
• Unevenness of moving of extremities during every moment of time;
• Constantly changing direction of a vector of the center of gravity;
• Emergence of uncontrollable external influences on the robot;
• A considerable deviation from a trajectory or falling in case of untimely or inexact reaction to external influences;
• Need of continuous correction of position of a body for saving of the steady provision.

Main purpose is development of algorithm which will provide stable walking of the robot. For achievement of a main purpose it is necessary to solve the following problems:

• To perform the literature review on a work subject;
• To make the analysis of existing methods of control over movement of walking robots;
• To develop model of movement of the robot which allows to describe precisely process of walking and the main impacts on it;
• To develop algorithm of movement which can consider all key parameters of walking, which can make balancing of position of a body of the robot at external influences;
• To develop the software for an assessment of efficiency of algorithm;
• To carry out testing of this software.

As a result of the carried-out review of literature the following approaches were allocated:

• Management of robot movement with use of static algorithms for ensuring stability of walking. For example in the RunBot robot as the stabilizer of gait «local reflexes» are used. It is the elastic reaction put in an extremity to loading. It is carried out return of an extremity to a starting position for maintenance of stable walking [6, 7, 11];
• Training to walking with use of genetic algorithms on a preliminary template. The approach means recalculation of operated parameters through the fixed temporary intervals. A lack of such approach is big computing loading on the robot and not optimized managements of walking because of not structuring phases of movement of extremities. Because of use of a template of gait, management of extremities is carried out in the small range allowed by a template. It is negatively reflected in result [1, 8, 9];
• Training by evolutionary methods with consecutive management of extremities. This approach doesn't demand considerable resources for calculations. Templates at this approach aren't used. However such approach can't provide high-grade walking. When walking during every moment of time all extremities should be used. In the absence of a template of gait and "not natural" consecutive control mode by extremities walking of such robot is absolutely not similar to walking of animals and its productivity is lowest [2, 3].

Picture 2.1 – Changing the vector of power acting on the robot during step

(animation: 7 frames, the delay between frames 0.5 sec, the number of cycles of reproduction – 6, size – 53 кб)

When developing algorithm of management of the walking robot it is offered to use best of features described above approaches. To refuse use of templates of gait and to exercise parallel administration of all extremities at the same time.

During the analysis of existing approaches, walking mechanisms as the environments most perspective and independent of conditions, a land relief, a soil covering were chosen.

In the field of walking mechanisms there are very big prospects. This area is the least studied. However it is widespread in wildlife. It represents considerable interest for a science [4].

When studying methods of management of walking mechanisms it was established, what the most perspective in the field of management of walking mechanisms are evolutionary algorithms. They have direct similarity with training methods to circulation in wildlife. Designs of robots often coincide on physical model with live organisms. Animals achieved tremendous successes in moving by such methods. It allows to use training methods from fauna. Also it gives the chance to estimate the received results on the basis of results of animals.

When writing this paper the master's work is not complete yet. Final end: December, 2012. The full text of work and materials on a subject can be obtained from the author or scientific adviser after this date.

1. Hein D. Simloid: Evolution of Biped Walking Using Physical Simulation / D. Hein – Berlin, Institut für Informatik, 2007. – 415 р.
2. Lipson H. Evolutionary Robotics for Legged Machines: From Simulation to Physical Reality / H. Lipson, J. Bongard, V. Zykov, and E. Malone // Intelligent Autonomous Systems 9 (IAS–9) – 2006. – P. 11–18.
3. Jakobi N. Evolutionary Robotics and the Radical Envelope–of–Noise Hypothesis / N. Jakobi // Adaptive behavior vol. 6 – 1997. – №9 – P. 325–368.
4. Zielinska T. Motion synthesis. Walking: biological and technological aspects / T. Zielinska – CISM Courses and Lectures 467 – 2004. –  P. 151– 187.
5. Hasimoto K.  Journal of the Robotics Society of Japan / K. Hasimoto, Y. Sugahara, M. Kawase, A. Hayashi, C. Tanaka, A. Ohta, T. Sawato, N. Endo, H. Lim,  A. Takanishi –  2007. –  №6. –  P. 53–60
6. Bai S. Terrain evaluation and its application to path planning for walking machines / S. Bai , K.H. Low // Advanced Robotics –  2001. –  №7 – P. 729–748.
7. Wongsuwarn H. Neuro–Fuzzy Algorithm for a Biped Robotic System / H. Wongsuwarn, D. Laowattana // Proceedings of the World Academy of Science, Engineering, and Technology – 2006. – №10. – P. 81–102.
8. Garder L. M. Robot gaits evolved by combining genetic algorithms and binary hill climbing / L. M. Garder, M. E. Hovin // Genetic and Evolutionary Computation Conference (GECCO 2006) – 2006. – vol. 2 Seattle – P. 1165–1170.
9. Palmer M. E. Evolved Neural Controllers for Bipedal Dynamic Walking with Multiple Demes and Progressive Fitness Functions / M. E. Palmer, D. B. Miller // Genetic and Evolutionary Computation Conference (GECCO 2009) – Montreal, Canada – 2009. – P. 289–301.
10. Kovalchuk A. The review of models of biped walking robots / Kovalchuk A. – Moscow.: MGOU, 2007. – 152 с.
11. Loffler К. Sensor and Control Design of a Dynamically Stable Biped Robot / К. Loffler, М. Gienger, F. Pfeiffer // Robotics and Automation. 2003: Proceeding. ICRA apos; 03. IEEE International Conference. W.t. –  2003. – vol. 1. Issue 14–19. – P. 484–490.