AAU Student Projects - visit Aalborg University's student projects portal
A master's thesis from Aalborg University
Book cover


One Step Closer; Towards Dynamic Walking on AAU-BOT1

Translated title

Et skridt på vejen, mod dynamisk gang med AAU-BOT1

Authors

;

Term

10. term

Publication year

2010

Submitted on

Pages

255

Abstract

Afhandlingen beskriver udviklingen af en dynamisk gangcontroller til AAU-BOT1, en tobenet robot fra Aalborg Universitet. Tobenede robotter er hybrider: størstedelen af gangen er kontinuerlige bevægelser, men hvert fodisæt er en diskret hændelse. Planlægning af en bevægelsesbane (trajektorie) og styring af robotten skal derfor håndtere begge dele. AAU-BOT1 kan ikke gå statisk på grund af kinematik og vægtfordeling samt små aktuatorer og stor friktion i leddene. Under udviklingen er det ofte foreslået at bruge en traditionel PID-regulator som servo og antage, at ikke-lineære effekter kan undertrykkes. Afhandlingen tester denne antagelse ved at tune og afprøve en PID-regulator på den fysiske robot. Gangen genereres løbende (online) via modelprediktion og numerisk integration af et ikke-lineært inverteret 3D-pendul, en standardmodel for balancerende gang. Resultaterne viser, at ikke-lineariteter kan opstå og destabilisere en PID-regulator. Ved at sænke tempoet i trajektorien og holde knæene strakte kan problemerne modvirkes, og der opnås stabil, quasi-statisk gang (langsomme skridt, hvor de dynamiske effekter er små, men ikke nul).

The thesis describes the development of a dynamic walking controller for AAU-BOT1, a biped robot from Aalborg University. Biped robots are hybrid systems: most of the gait is continuous motion, but each foot impact is a discrete event. Planning a trajectory and controlling the robot must account for both. AAU-BOT1 cannot walk statically because of its kinematics and weight distribution, combined with small actuators and high joint friction. During development, it has often been suggested to use a traditional PID controller as a servo and to suppress non-linear effects. This thesis tests that assumption by tuning and applying a PID controller on the physical robot. The gait is generated online using model prediction and numerical integration of a nonlinear inverted 3D pendulum, a standard model for balancing during walking. Results show that nonlinearities can arise and destabilize a PID controller. By slowing the trajectory and keeping the knees straight, these issues are mitigated, and a stable, quasi-static gait is achieved (slow steps where dynamic effects are small but not zero).

[This abstract was generated with the help of AI]