2013 Student projects:

Group 13gr1034 from Section of Control and Automation:
Approaching Dynamic Gait with AAU-BOT1

Niels Hyltoft Andersen, Rune Madsen

Abstract

This thesis concerns the topic of obtaining dynamic gait with the AAU-BOT1. Previous work on the AAU-BOT1 has resulted in a functional robot, where each joint can be individually controlled. Themain topic has therefore been to design and implement a controller strategy capable of obtaining dynamic gait with the robot. Through a study of the human gait, including recording human gait in a motion tracking room, knowledge about human gait has been obtained. Kinematic and dynamic models are derived for the robot. From these the overall movements to obtain gait is deduced. Two levels of controllers are used to control the robot. Low level controllers are used to control each separate joint. A high level gait and balance controller is used to generate joint references and keep the robot in balance. The perception of balance is gained by measuring the angle of the torso and the interaction between foot and ground. The foot ground impact is given special attention. A shock absorption strategy is used to reduce harmful effects of hard impacts. With the implemented controller strategy the AAU-BOT1 has been shown to be capable of obtaining gait with a velocity up to 0.15 km/h.

Trainee student from Eindhoven University of Technology, Holland:
Modeling for control of a biped robot - AAU-BOT1
Another step closer towards dynamic walking

Bart Moris

Abstract

The focus during this internship has been to investigate the existing dynamic model of the robot, to improve it if necessary and to do research on joint trajectories. This work is fully based on simulations with an existing model derived by previous project work, and this model has been point of origin of the current project. Particularly investigating the implementation of the dynamic model of AAU-BOT1 and improving the dynamic model has been a major part of the research carried out in the last three months. This has resulted in the identification of several significant errors in the dynamic model limiting the ability to perform simulations giving appropriate results. This report includes moreover an enclosed CD with the most recent simulation data and the updated library as described in Appendix E. To me, the last three months have been a wonderful and moreover instructive and useful contribution to my Master studies, where it was especially exiting to participate in this big project. I hope that my contribution is useful in the future and that within a reasonable amount of time the final goal is achieved and AAU-BOT1 will walk.

2010 Student projects:

Group 10gr1036 from Section of Automation and Control:
One Step Closer; Towards Dynamic Walking on AAU-BOT1

Jes Andersen, Henrik Dalsager Christensen

Abstract

The thesis treats the development of a dynamic walk controller for AAUBOT1. AAU-BOT1 is a biped robot developed at Aalborg University. A Biped Robot system, can be described as a hybrid system, as long time spans of the gait are governed by continous motions, and the impacts with the ground can be considered discrete events. Planning a trajectory, and applying it on the system must take this into account. AAU-BOT1 is physically unable to walk statically, this is partly due to kinematics, and weight distribution, and partly due to the combination of small actuators and large frictions in the joints. It has been suggested numerous times during the development, that a traditional PID regulator can be used as a servo controller, and that all the nonlinear effects, both discrete as well as continuous can be suppressed on the robot. This claim is put to the test in this thesis, as a traditional PID regulator is tuned and applied for the physical system. The gait is generated on line, based on model prediction, and numeric integration, of a nonlinear inverted 3d pendulum. Applying the trajectory to the robot requires that the implemented PID controller can handle the system suficiently well, it is however shown that non-linearity's can occur which can destabilize a PID regulator. Slowing down the trajectory, and keeping the knees stretched counters the problems, and a stable quasi-static gait is achieved.

2009 Student projects:

Group 09gr1031 from Section of Automation and Control: Modeling, Simulation, and Control of Biped Robot AAU-BOT1

Brian Thorarins Jensen, Michael Odgaard Kuch Niss

Abstract

The aim of this masterís thesis is to further development of AAU-BOT1, which is a bipedal robot with human proportions, designed and built at Aalborg University. The robot is designed to combine the fields of robotics and health research in the study of gait patterns including dysfunctional limbs. To provide a platform for this project and further development on AAU-BOT1, a complete hardware and software platform is set up to allow for easy implementation of new control systems. A complete model of the robot is set up in order to simulate the robot realistically and to be used in the control system design. To make the robot walk static balanced gait a trajectory is generated which meets the physical limitations of AAU-BOT1. To follow the trajectory a control system is created based on an unscented Kalman estimator, a posture controller, and a balance controller. The estimator is designed to obtain an estimate of unmeasured system states needed in the posture controller, which consists of feedback linearization and decoupling based on the computed-torque control method, combined with linear controllers. The control system is tested with changes in the parameters of the system model to reflect model uncertainties compared to the real robot. The final result is a complete development platform for AAU-BOT1 and a control system which can make AAU-BOT1 perform static balanced gait in simulations.

2008 Student projects:

Group 08gr1032b from Section of Automation and Control: Instrumentation, Modeling and Control of AAU-BOT1

Per Kingo Jensen, Mathias Garbus, Jan Vestergaard Knudsen

Abstract

The aim of this masterís thesis is to equip the humanoid robot AAU-BOT1 with sensors, model and control it such that it can obtain static gait. AAU-BOT1 has human proportions and features 17 actuated degrees of freedom. To enable AAU-BOT1 to obtain static gait, an instrumentation strategy has been pro- posed and implemented. Furthermore a software platform is developed to complete the instrumentation. Models of the DC-motors, kinematics, inverse kinematics and the dynamics of AAU-BOT1 have been made. By utilizing the inverse kine- matic model, static gait trajectories are developed. The remaining models are utilized to create two different control strategies. The first control strategy is based on a Linear Quadratic Gaussian controller(LQG), which controls the posture of the robot based on the dynamic model. The second control strategy is based on classical PID controllers, and utilizes the build in features of the digital DC motor amplifiers. Both control strategies contains a balance controller, that is used to maintain stability during walk. It is furthermore decided to develop a virtual representation of the AAU-BOT1 for test purposes. The LQG controller strategy with the proposed trajectories was tested on the virtual AAU-BOT1 but the controller could not stabilize the robot sufficiently. The second control strategy was successful and the virtual robot is able to walk with the proposed trajectories and maintain stability at the same time. The second control strategy was only partially implemented on the physical AAUBOT1 and showed promising results of obtaining static walk.

2007 Student projects:

Group 43A from Department of Mechanical Engineering:
Design of Robot AAU-BOT1

Mikkel Melters Pedersen, Allan Agerbo Nielsen, Lars Fuglsang Christiansen

Abstract

The aim of this project is to develop the mechanical design, insluding power transmission, of a dynamic walking humanoid robot.

Initially existing robots and human walking principles are investigated, and the requirements are specified.

The dimensioning loads are estimated through inverse dynamic analysis af recorded human motion from experiments.

Suitable actuators are selected, and the mechanical design is developed to accommodate these. The design includes a lightweight six-axis force/torque sensor which shall provide input for the final control of the robot.

The final design and composition of actuators are verified by time domain simulation. Through the implementation of a preliminary control strategy, it is concluded that dynamic walking is possible.

Lastly, an optimization scheme for weight minimization of the structural parts is set up, and applied to the heaviest part of the robot.



Group 1033 from Section of Automation and Control:
Development,modeling and control of A Humanoid Robot

Jens Christensen, Jesper L. Nielsen, Mads S. Svendsen, Peter F. ōrts

Abstract

The robot is designed to resemble human proportions and a special joint has been developed to resemble the hip joint of humans, and thereby enabling walking in curved paths. Furthermore the hardware necessary to obtain a fully automomous system is developed and implemented. The result of the design phase is a humanoid robot, called "Roberto", measuring 58 cm and with 21 actuated degree of freedom. A complete dynamical model describing the system has been developed. The model is a hybrid model which enables siimulation of complete walking cycles. A novel solution of the dynamics of the robot during double support phase has been given.

To enable human-like walk a set of trajectoties has been developed, based on the zero-moment point and simulations. The trajectories are simulated, and human-like walk is obtained on the model. To maintain stability during walk with the real robot, two controllers have been developed, a posture controller and a zero-moment point controller. It was found that the controllers were able to track a zero-moment point breference and a inclination reference given to the system.

Human-like walk was not obtained due to system limitations. New servos with higher up date rate is necessary.