The dissertation presents a novel approach of an actuation system design for bipedal robotic walkers. The new actuation system concept is, to a certain extent, a biomimetic solution based on biological walkers. It consists of using elastic cables to actuate a bipedal walking robot, where the motors, which are used for pulling the cables, are attached to the segments that undergo low accelerations (like trunk).
The main results and contributions of the dissertation are:
(a) The new actuation system has better acceleration capabilities of the walking robot, when compared to the conventional actuation system, which is based on joint attached motors,
(b) The new actuation system consumes less energy due to reduced inertia of the most kinetically active links of the walking robot,
(c) Using the new actuation system, the walking robot architecture is compact compared to conventional walking robot architecture,
(d) Biped robot has better stability (margins) properties, due to the reduced perturbations associated with inertial effects (lower inertial forces and moments),
(e) A novel robotic combined system of bipedal walking robot with a rotor was designed, which is capable of walking and hovering.
Particularly, a bipedal walking robot with 21 degrees-of-freedom (DoF) was designed with emphasis on the way the robot is actuated. Traditionally, walking robots are actuated using joint directly attached rotary motors. Since most of walking robot mass is associated with the mass of motors, the additional mass at the robot joints causes significant disturbances throughout the walking cycle due to the high inertial effects.
Particularly high accelerations can be sensed at the distal segments of the robot, such as lower legs, knees, and ankles throughout the swing phase of the walk, resulting in high inertial forces. Frequent acceleration and deceleration of the joint attached mass costs a significant amount of energy per each cycle.
The novel design of walking robot actuation via elastic cables has two implications on the robot design and performance. It significantly reduces the inertial effects by dislocating the motors from the most kinetically active robot segments, experiencing high acceleration magnitudes, to segments with lower accelerations, such as the trunk, and transferring the motor torques to the joints via cables. The second important benefit is that joints with multiple DoF, such as ankles and hips, can be designed and actuated in a similar way as in biological systems (e.g., hip joint is a ball-and-socket 3 DoF joint), rather than representing multiple DoF joints as multiple revolute joints separated with additional links.
A cable-based actuator design would require n+1 cables per each n-DoF joint (e.g., a revolute joint would require 2 cables), with n+1 motor to pull the cables independently in case of inelastic cables. Using elastic cables instead of inelastic cables, two main goals are achieved: the number of motors needed for pulling the cables is reduced and the robot is enabled to smooth out the foot-ground impact forces and corresponding effects by accumulating the impact energy and releasing it throughout the walking cycle.
As a part of this dissertation, a specific architecture and application of biped walking robot has been designed. The design is specific in the sense that it is a combination consisting of a helicopter-like rotor with the dual-purpose arms shaped as propeller blades and a cable-actuated biped taking the role of the rest of the architecture. The main purpose of such a combined robot is to enable the robot to hover and reach disaster-hit areas and perform search-n-rescue operations.
A model-based controller for the bipedal and quadrupedal walk is designed and tested throughout simulations, which are presented in the dissertation with accompanying discussion and recommendations for further work and improvements. Hardware implementations of a quadrupedal and a bipedal walking robot with tests results and analysis are presented.