A lightweight wall climbing robot with reduced degrees of freedom (DOF) is under development. This robot consists of three main components. The first component is a quadruped walking robot with reduced-DOF. Since lightweight is an important feature for wall-climbing robots, the DOF of this robot were reduced until three, without hampering its ability as a walking robot. The second one is a sucker using a blower. The adhesion force with a blower is weaker than with a vacuum pump, but the blower can deal better with air leakage. The sucker developed in this research can adhere to walls with many slits and uneven ceiling. Finally, the last component is an ankle joint with wide movable range. This mechanism is indispensable for walking freely in any environment and transferring from wall to wall, wall to ceiling and ground to wall. Moreover, the appropriate compensated motion for wall climbing is considered. The structure of the wall climbing robot has been deformed by the gravity effects. Therefore, the compensated motion should be indispensable for its smooth climbing motion. Wall climbing experiments using the walker developed in this research have been performed with the proposed compensated motion

This paper proposes a new mechanism and control method of 4-leg-wheel hybrid locomotion. Many previous hybrid vehicles execute wheel-mode and leg-mode alternatively. However, to overcome higher obstacle with sufficient velocity, a wheel locomotion should continuously be executed. In this paper, suitable load distribution control is proposed considering a contact force to the ground and step edge, a wheel driving torque, and a friction coefficient of the wheel and terrain. In a proposed new leg-wheel hybrid mechanism, load distribution control is realized by a torque of twisting joint at the center of the body, and a forward/backward shift of the body. The validity of this control and mechanism is confirmed by some experiments, where vehicle can run over a 105 mm step by 90 mm radius wheels with a good continuous movement. With an adjustment control of the approach angle, a robot can also ride over a 230 mm step

From 1970's, legged robots have attracted much attention of many researchers. In spite of this, it has been regarded that dynamically stable walking is very difficult to be tackled for any types of legged robots. For a trot gait for quadruped walking robots, we have proposed "the sway compensation trajectory". This method utilizes a lateral, longitudinal, and vertical motion of a robot body to keep a zero moment point (ZMP) on a diagonal line between support legs. In this paper, we develop the sway compensation trajectory for a biped robot, and show that dynamically stable walking is realized. This method makes it quite easy to design stable ZMP and COG (center of gravity) trajectories, which have been regarded as a very complicated and delicate problem. The effectiveness of the proposed method is verified through computer simulations and walking experiments by a humanoid robot, HOAP-1.

At present, robots are demanded not only for stationary use but also for tasks that require high mobility. The important abilities for such robots are both moving and task performing abilities. However, in order to perform tasks and move using separate mechanisms, robots are required to have many degrees of freedom (DOF). As a result, robots become heavy and lose mobility. In this paper, an effective design of robots with reduced-DOF is described. Besides, an 8-DOF bipedal configuration robot with high tasking and moving performance was developed. In this robot, the DOF have been reduced as much as possible without hampering its mobility, finding a balance between the increase in weight with additional DOF and the loss of manipulability with reduction of DOF. Moreover, applying this concept to the locomotion using the same DOF for differential purpose, it is possible to use the sane mechanism for the legged locomotion and wheeled locomotion. In this paper, this concept and maneuvers for locomotion are described. New experimental model is designed and experiment results for proposed locomotive maneuvers are expressed.

In order to perform tasks and move using separate mechanisms, robots are required to have many degrees of freedom (DOF). If robots can perform both wheeled locomotion on flat ground and legged locomotion on uneven terrain, they can achieve much higher mobility. We have already proposed an effective design method with both mobility and manipulability, that is a method of combining each mechanism. The designed robot uses the same mechanism for locomotion and manipulation. By applying this concept to locomotion, it is possible to use the same mechanism for legged locomotion and wheeled locomotion. In this paper, this concept and maneuvers for locomotion are described and an experimental model and experiment results are reported.

Some stability criteria for walking vehicles on rough terrain are discussed. Several criteria for stability margin have been proposed previously and they can be roughly divided into three categories. The paper compares them with each other and concludes that a stability criterion based on energy consideration is the most reasonable for practical use through simple experiments. The existing stability margin is apparently inadequate because it varies with weight of the vehicle with the same posture, while weight doesn't affect its resistance to a tumble. Therefore, an improved stability margin normalized by weight is proposed. At the same time, a helpful tool to derive the desirable posture for walking vehicles is described which can maximize the proposed stability margin on rough terrain.

We propose a new trajectory planning for the stable trot gait, named 3D sway compensation trajectory, and show that this trajectory has lower energy consumption than the conventional sway trajectory proposed previously by Yoneda et al. (1995). As the adaptive attitude control method was used during the 2-leg supporting phase of the trot gait, we consider four methods: 1) rotation of body along the diagonal line between supporting feet; 2) translation of body along the perpendicular line between supporting feet; 3) vertical swing motion of recovering legs; and 4) horizontal swing motion of recovering legs. The stabilization efficiency of each method was verified through computer simulation and the damping experiment using a quadruped walking robot, TITAN-VIII. Furthermore, the dynamic trot gait control that combines the feedforward control based on the proposed 3D sway compensation trajectory and the adaptive feedback control using body translation and vertical motion of swing legs was developed, and the walking experiment on rough terrain using TITAN-VIII was carried out.

In order to perform tasks and move using separate mechanisms, robots are required to have many degrees of freedom (DOF). As a result, robots become heavy and lose their mobility. In this paper, an effective design of robots with reduced DOF is described. An 8-DOF bipedal configuration robot with high tasking and moving performance was developed. In this robot, the DOF are reduced as much as possible without hampering the its mobility, by finding a balance between the increase in weight with additional DOF and the loss of manipulability with a reduction of DOF.

To introduce the force control for an uneven terrain walking robot such that it effectively sets the suitable contact pressure of its supporting legs and is able to change from support to swing legs smoothly, the force control for ParaWalker-II, serial/parallel hybrid mechanism that is connected between two frames and contacts ground at three points, is proposed. With this method, the entire leg-drive mechanism acts as a 6-axis force sensor; and the force and moment can be calculated from the changes of each joint angle such that sensors at the ends of the leg are not necessary. Furthermore in the paper; force control and position control can be chosen for each direction in order to smoothly walk on uneven terrain. Smooth contact with the ground by the swing-leg and the changing from swing to support leg is verified through experiments using this force control and measurement method.

A walking robot that has high ground adaptability and the ability to perform various tasks would be extremely desirable in many fields. It can move efficiently in an outdoor environment and on uneven terrain such as stairs, as well as perform tasks. In this paper, for a six-legged uneven terrain walking robot with two body segments, we examine the number of degrees of freedom, movement performance, stability on uneven terrain, and work performance. In addition, we examine the arrangement of the degrees of freedom suitable for practical uneven terrain work by this type of robot. We also propose that a walking robot with an extra leg extension mechanism and a 6-DOF between the body segments is effective. We adopt this design to ParaWalker-II, which we had previously developed, and verify the effectiveness.

A new lightweight six-legged robot is developed which uses a simple mechanism and can move and work with high efficiency. This robot consists of two leg-bases with three legs each and walks by moving each leg-base alternately. These leg-bases are connected to each other through a 6-DOF mechanism. Thus, when the robot stands on three legs, or one leg-base is placed on the ground, the other leg-base can be used as a 6-DOF manipulator or active working platform. The output force, velocity, and movable range of various mechanisms for connecting the two leg-bases were compared and the results showed that good performance could be achieved with a serial-parallel hybrid mechanism consisting of three 6-DOF serial linked arms positioned with radial symmetry about the center of each leg-base, each composed of two active and four passive joints. Walking experiments with this robot confirm that this mechanism has satisfactory performance not only as a walking robot but also as an active walking platform.

Design and control of a new quadruped walking and manipulating robot TITAN VII which can be operated on a steep slope is discussed. We first consider the optimal design of a driving system of legs consisting of prismatic actuators so as to produce large output power with minimum weight based on the concept of coupled drive. Secondly, we discuss the criteria for stability on a slope and introduce a new type of extensible leg to maintain high terrain adaptability on a sleep slope. Thirdly, we mention a passive terrain adaptive foot mechanism consisting of a "rocker bogie sole" based on the concept of a connected differential mechanism and including an ankle installed sensor mechanism. The validity of these newly introduced mechanisms has been investigated through walking experiments of the prototype model TITAN VII, including climbing a slope of about 30 degrees using the intermittent crawl gait.

This paper discusses an algorithm to make a quadruped machine walk dynamically and omnidirectionally following a real-time command from an operator. To produce a smooth body motion at high speed locomotion, we introduced a new gait, named the "intermittent trot gait", in which a four-leg-supporting phase and a two-diagonal-leg-supporting phase appear in orders. In this gait, the pitching and rolling motion of the body is suppressed to a lower level than that of the other gaits because the dynamic effects of the swinging legs almost cancel each other. To realize the intermittent trot gait, new algorithms to plan a landing point and body motion are discussed. A landing point is decided by considering an operator command and conversion to the standard leg formation. A body motion is planned to produce dynamic stability and to follow speed and direction commands. By using these algorithms, a mechanical vehicle TITAN VI can walk omnidirectionally, smoothly following a real-time operator command. The top speed is 1 m/sec on flat terrain and 125 mm/sec on 105 mm up and down over rough terrain.

On the assumption that a quadruped robot works on a slope, we discuss how to make it prevent tumbling over. The larger the difference becomes between the potential energy of the center of gravity of the initial position and that of the highest position after its rotating, the less the robot tumbles. So this difference can be regarded as stability margin, and a novel gait to obtain largest stability margin is mentioned here. It is an intermittent crawl gait. Its energy stability contour (consisting of equal stability points on the inclined plane) is helpful in the design of standard foot trajectories. An optimal posture on the slope, designed in this way, results in inverse trapezoid shape, which means that upper two legs are located wider than lower two ones. This form worked for the experimental machine, TITAN VII. Furthermore, if the standard trajectory for one direction is combined with another direction trajectory, the quadruped robot can easily switch its proceeding directions, keeping enough stability margin. This sequence is shown.

We discuss methods to evaluate the stability of robots which execute manipulations and locomotion. The relationship between the center of gravity projection point and the ground contact points is an easy-to-use criterion but it does not take into consideration manipulation counter-force. We propose a concept called "tumble stability", which considers the tumble direction when all ground contact points except 2 are hypothetically lost. If a given, hypothetically lost, ground contact point can produce support force which can suppress that tumble, the walking machine will not tumble; otherwise, it will. Stability during manipulations can be evaluated by incorporating the manipulation counter-force as an external force in this evaluation standard. The stability of the planned manipulation can be determined not only on level ground, but on general topography including rugged terrain, wall surfaces and ceiling surfaces. We introduce the concept of the manipulation counter-force limit as an index to indicate manipulation capabilities. Because this indicates the limit value of the counter-force by which a manipulation can be stably executed, it clarifies the fact that it is effective to make the ground contact points on the side opposite from the manipulation farther and wider in order to heighten manipulation performance.

Outlines the fundamental considerations for a planetary rover. First, shapes which utilize limitless rotation such as wheels and crawlers, legged shapes, and articulated body shapes were compared, and it was found that the wheel type is currently the optimum for a planetary rover. Next, the authors study specific methods for configuring planetary rovers with one, two, and three wheels, which is a topic that has not been studied much before, and in particular, the authors indicate that it is effective to equip these vehicles with manipulator arms and make joint use of the arm for movement. The authors also study a four-wheel rover, and propose a rhomboid four wheels vehicle arrangement as a mechanism to manifest terrain adaptability comparative to the six wheels vehicle.

This paper discusses foot force control of a quadruped walking vehicle on rough terrain. The proposed control algorithm has two parts. One is a feedforward control to produce a smooth touch-down and lift-up motion of the leg. The other part is a feedback control to maintain the body's posture under the conditions of some disturbances. A command force of the feedforward control is calculated beforehand from the data of the body's position, its acceleration and leg positions. The force increases smoothly from zero after touch-down on the ground, and it decreases smoothly to zero before lift-up. On the other hand, a command force of the feedback control is calculated from the data of the body's posture measured by gyroscopes. The effect of the feedback control is as if the body is suspended to an absolute frame by springs and dampers. Therefore, it is called 'Sky-Hook Suspension'. By this control-algorithm the experimental vehicle TITAN VI is able to walk smoothly on a rough terrain.>

This paper describes a solution to the problem of planning a leg motion for a six legged robot when smooth versatile motion commands of its body are given. The well-known wave gait for a longitudinally symmetric robot is generalized to solve the omnidirectional gait planning problem for a radially symmetric six legged robot. The algorithm can handle any combination of translational and rotational motions for a variable crab angle condition as long as the rates of speed and crab angle are moderate. Guaranteeing the maximum duty factor for any given commanded motion, this solution optimizes the stability margin of the robot. The foot trajectory planning algorithm dynamically generates a smooth foot trajectory as a function of the instantaneous body motions.>

The authors previously (1989) proposed the concept of dynamic and static fusion gait control and discussed a basic control scheme for the purpose of adaptive selection of these gaits during continuous walking. They extend the concept to walking at variable speeds along a winding path. The motion scheme is discussed on the basis of a generalized trot gait and incorporates stability control based on feedforward and feedback control; trajectory modification is based on the zero moment point concept. The validity of these discussions is verified by walking experiments with quadruped walking vehicles TITAN IV and TITAN VI; e.g., dynamic and static fusion gaits at velocities from zero to 400 mm/sec by TITAN IV, and walk with dynamic stability along a circular winding path at a velocity of 100 mm/sec by TITAN VI. At present, dynamic walk on rugged terrain is not possible, but it may be made possible by introducing feedback control as discussed.>

A six-axial force sensor using the optical measuring technique and its nonlinear calibration method are proposed. The force sensor is based on a unit which has a small light source set face to face with a photosensor of the quarter-splitting type to measure minute displacements in two directions with respect to each other. Three sets of the unit are held by an elastic frame. The sensor, in comparison with the conventional strain-gauge-based device, is more compact, light in weight, low in cost, and accurate. The high accuracy of the sensor comes from the calibration method in which nonlinear interferences of six-axial force are considered. An experimental setup in which the six-axial force sensor could be simultaneously input with plural axial forces was produced, and the proposed calibration method was shown to be valid.>