Objectives & Methodology
The development of biomimetic robots that can operate in rough environments demands an interdisciplinary approach. To develop robots that can close the performance gap with their biological counterparts, a number of important research objectives must be addressed. The research program will build upon the past work and experience of the three teams, and will include analytical, computational, and experimental components. The main program objectives and the methodology the will be followed are outlined below.
Analysis of human and animal locomotion. The locomotion mechanism that is used by humans and animals to transverse rough terrains and difficult environments are not well known yet. The UTH team will conduct three-dimensional kinematic analyses of humans and animals, e.g. of felids or canids, using the Vicon's motion capture system combined with measurement of the ground reaction forces. The individuals will walk or run at different speeds on flat and uneven tracks with a perturbation consisting of a force plate varying in height. From the measured kinematics and kinetics we will calculate the body potential and kinetic energy. Moreover we are going to investigate how participants change their gait characteristics and especially the leg behavior to sustain stability.
Development of legged robot models. The NTUA team will start with forming a coupled three- dimensional model of quadruped robot motion in the transverse and longitudinal planes. Initially, 2-dof legs with actuated hip joints and unactuated prismatic springy joints will be assumed. The addition of a minimum of joints to the body and the legs will be studied and new models will be developed. It is expected that the addition of a transverse rotational DOF to the robot main body will allow for higher speeds and that one rotational body DOF with its axis parallel to the robot’s longitudinal axis will lead to a more stable, three- dimensional locomotion, e.g. asymmetric galloping. In addition, legs with variable compliance, as a function of ground stiffness and robot dynamics, will be modelled. It is expected that these models will lead to the design of multi-joint, variable compliance legs that will improve handling of slopes and terrains with obstacles. Alternative gaits will be analysed, and the conditions for successful transitions between gaits will be investigated using nonlinear dynamics. Models will be evaluated by simulation using a Mathematica computational testbed based on Lagrangian dynamics. For simulations, ADAMS and Matlab will be used. The FORTH team will develop analytical and computational models for segmented multi-legged undulatory and pedundulatory for biped humanoid robots, moving on unstructured environments, and perform parametric and model validation studies. During the development of the quadruped, multi-legged and humanoid models, results from animal and human locomotion analysis will be taken into consideration.
Stability investigation and development of control algorithms. The NTUA team will develop systematic, model-based control methods that exploit the inner dynamics with emphasis on the control of quadruped robots with multi-joint, variable compliance legs and articulated, compliant body, capable of changing motion parameters impromptu. This area of research will be studied in conjunction with the dynamics of the system and the environmental conditions, e.g. slopes, stiffness, etc. The objective is to design robust controllers that guarantee system stability in the presence of small disturbances, and that allow setting desired goals, such as robot speed, apex height, gait transition, level of compliance. To this end, tools from the areas of nonlinear dynamics (nonlinear oscillators, limit cycle analysis, Poincaré maps) and nonlinear control (underactuated, hybrid, backstepping, and robust model-based control) will be employed. Controllers will be model based and will be evaluated by simulation using the MATLAB and ADAMS software packages. It is expected that this methodology will set the base for the development of a more efficient robot, both because less energy is required, and because the overall system mass is reduced. The FORTH team will exploit the developed computational models of multi-legged and humanoid robots, for the design of gait generation and of sensor-based closed-loop motion control for locomotion in unstructured environments.
Design and development of novel robotic platforms. Based on the obtained results, the NTUA team will design and build an experimental quadruped robot with multi-joint, variable compliance legs and articulated, compliant body to to conduct experiments for the evaluation of the theoretical analysis. During the development of the quadruped robot, results from animal and human locomotion analysis, as well as those from multi-legged and biped robots research, will be taken into consideration. It is expected that the new quadruped robot will perform dynamic stable motion at high speeds through uneven terrains. The FORTH team will develop multi-legged robotic prototypes including novel actuation methods, flexible elements, and implementing a distributed control and sensing architecture. Studying the interaction between these robots and different substrates, on custom test-beds, will reveal how the former can handle terrain irregularities.
Experimental evaluation and redesign. The UTH team will conduct walking and running experiments with humans, animals and developed robotic systems to evaluate robot performance and the correlation between machine and animal locomotion. The NTUA team will conduct experiments with the novel quadruped robot to evaluate and validate the theoretical analysis. During this process results will be used for redesign. The FORTH team will conduct experiments with its multi-legged robotic prototypes, for evaluation and, possibly, model/control/prototype enhancement or redesign. The FORTH team will also use its FUJITSU HOAP-3 robot, or other analogous robots, to evaluate its bipedal locomotion control strategies. The results will be compared to those obtained by UTH for human locomotion analysis.