This paper discusses the refinement of sparse and noisy depth-maps to improve stereo measurements. Our method functions as a post-filter for stereo measurements, to remove outliers and interpolate the depths of invalid pixels. Per-pixel plane fitting is employed to estimate the normals of an object's surface in a depth-map. These normals provide information regarding the interpolation of depth and the removal of outliers by evaluating the directions of surfaces. In our experiments, our method successfully reconstructed a dense and accurate geometry from a sparse and noisy depth-map, even where several dozen percent of pixels were outliers and only a few percent were from the original correct geometry. This result indicates a novel method of fast stereo measurement, because dense reconstruction can be performed without stereo matching for all pixels.

To realize a new type of personal vehicle for elderly people while contributing to a low-carbon traffic society, we have been developing a robotic wheelchair with autonomous traveling and obstacle avoidance in an urban environment. In this paper, we primarily discuss two kinds of key technologies, long-distance autonomous travel in an outdoor environment and obstacle avoidance in a human-shared environment. For the localization and navigation methods, we propose sub-map dividing and realignment with FastSLAM, which enables generation of large-scale 3D voxel maps by the sampling based SLAM method. For the planning and obstacle avoidance methods, we propose motion control by a combination of the global/local A* algorithm and the dynamic window approach. To confirm the effectiveness of our proposed methods, the results of demonstration experiments using our robotic wheelchair ‘Marcus’ in the Tsukuba Robot Zone are also reported.

We are developing a human-riding wheeled inverted pendulum for use as a personal vehicle. To be practically usable, the vehicle requires a function that enables the rider to get on and off both safely and smoothly while on a slope of unknown angle. Concretely, for the convenience of the rider, the vehicle needs to be stabilized on an unknown slope without the rider aboard, both before the rider gets on it and after he/she gets off it. Moreover, the vehicle should be stabilized safely relative to the rider's handling force at a grip of the vehicle while getting on and off. We thus developed a method of estimating the handling force and the slope angle separately. In this paper, we report the method to estimate the handling force and the slope angle using a disturbance observer. We verified the validity of our proposed estimation method by computer simulations and experiments using a prototype of a human-riding wheeled inverted pendulum on an unknown slope.

This paper reports on the development of the "TAO Aicle", an intelligent wheelchair with autonomous traveling capability. The wheelchair can accurately determine its position and direction using internal sensors and external assistance from the surrounding environment, including the Global Positioning System (GPS) and Radio Frequency Identification (RFID). It can also detect obstacles it encounters on the road using a laser range sensor. To confirm the effectiveness of our research results, for six months at the Aichi Expo 2005 we demonstrated the wheelchair system by offering trial rides to visitors. This paper describes the total system of "TAO Aicle", the positioning method using GPS and RFID, and the autonomous traveling control based on accurate estimations of position and direction using Kalman filtering. Our experimental results at the Aichi Expo demonstrated that accurate trajectory following performance can be achieved using assistance from the infrastructure (GPS, RFID) and that "TAO Aicle" could perform safe autonomous traveling without severe accidents.

This paper describes an automated device sizing system for current-steering D/A converters (DACs) and an 11-bit 160-MS/s DAC implemented using this system. Based on an analysis of harmonic distortion (or spurious) of the DAC, a circuit technique named One-Vgs Switching has been newly developed for realizing high spurious free dynamic range (SFDR). The automated device sizing system has also been developed for quick retargeting of the current-steering DAC. The 11-bit 160-MS/s DAC has been designed using this system and fabricated in a 0.18-μm technology. It operates at 1.35-V power supply with 10-mW power consumption, 1.6-Vppd output swing, and 61-dB SFDR at fsig=10.2 MHz. Its active area is 0.22 mm2.

Nonstop update (NSU) is a service robot architecture to upgrade running robot controllers without suspending the services of robot systems. Its target application is mobile robots operating in remote or hazardous environments where suspension of ongoing missions is unacceptable or human intervention for changing or debugging programs is not easy. The present paper introduces NSU from a point of view of both safety and performance. The present paper defines the safety in NSU as a property that controller programs should not suspend or run away because of update-related sources. Safety issues such as deadlocks, module dependency, discontinuity of behavior between before and after update, and delays due to update overhead time are addressed. Safety and performance of NSU is examined and demonstrated by using three practical mobile systems: a mobile manipulator for a power assist system, a hexapod legged mobile robot, and a biped walking robot with wheeled legs.

This paper reports on the continued development of a new concept for a personal vehicle called the 'personal riding-type wheeled mobile platform (PMP)' that consists of two wheels and a standing base for a human rider. The two wheels are driven independently, and forward and backward movement and steering are achieved by simply changing the relative position of the rider's center of gravity (COG) on the base. The vehicle has two distinct advantages: a reduction in total weight through its simple structure and a space-saving design that does not use a steering unit. We have already proposed the first prototype (PMP-1) and its posture stabilizing and running control methods, and we have achieved proper forward and backward movement according to the rider's intentions. This paper describes our proposed structure for detecting the rider's COG on the base and the steering control method in order to achieve steering control by changing the position of the rider's COG and the second prototype (PMP-2) whose weight is smaller than 12kg. We also investigated the steering control method to improve maneuverability in various estimated standing poses. Our experimental results demonstrate that natural steering control can be achieved using the rider's COG based on the rider's intentions.

We propose a new concept in personal vehicles called the 'Personal riding-type wheeled Mobile Platform (PMP)' consisting of two wheels and a standing base for a human. The two wheels are driven independently, and traveling and steering is achieved by simply changing the relative position of the rider's center of gravity (COG) on the base. The vehicle has two merits: reducing the total weight of the simple structure and saving space by not using a steering unit. In this paper, the mechanism of the first prototype, PMP-1, is introduced and its posture stabilizing and running control methods are proposed. The effectiveness of the posture stabilization control methods is demonstrated in computer simulations and human riding experiments. In the running control method, maneuverability by movement of the rider's COG is also investigated experimentally. The results demonstrate that the posture of the rider can be maintained and the velocity of forward and backward traveling can be controlled according to the rider's intentions.

This paper introduces a biped robot named TOKAI ROBO-HABILIS 1. It has some particular mechanisms in the joints, by which it moves with the combination of active driving by the DC motors and passive rotation of the joints, called active/passive hybrid walking. It is essentially the identical way of walking to those of human. As a result, it walks with a subsequently small energy consumption compared to any conventional biped walking robots that walk with driving all of joints actively by actuators. This paper describes at first the mechanisms of TOKAI ROBO-HABILIS 1. Its computer simulation shows the control strategy and the energy consumption for various walking patterns. It follows to show the walking experiments.

Combination of wheels and legs has the potential to realize efficient locomotion on a flat surface and also negotiation of artificial irregularity such as stairs or steps. We have been studying on biped leg-wheeled system to realize flexible locomotion in our daily, life environment. This paper describes the design concept of the self- contained biped leg-wheeled system and its locomotion concept. The system has three locomotive modes - 4-wheeled locomotive mode, 2-wheeled locomotive mode, stair negotiation mode. In the first mode, it can travel quickly and steer like a car with its four wheels contacting the ground In the second mode, it can travel and steer like a wheeled inverted pendulum with two wheels contacting the ground In the stair negotiation mode, it can negotiate the stairs with front and rear wheels continuously in contact with the horizontal and vertical surfaces of stairs. In this mode torque control of both wheels while maintaining the posture of the system is the key to successful negotiation. The control scheme and method for stair negotiation, which can be applied generally to a leg-wheeled system, are introduced The effectiveness of the system is demonstrated by a successful experiment of the lift-up motion of the wheel using a biped leg-wheeled system developed in the authors' previous research.

Meltran V, a new biped robot with telescopic legs, is introduced. For 3D walking control of the robot we analyze the dynamics of a three-dimensional inverted pendulum in which motion is constrained to move along an arbitrarily defined plane. From this analysis we obtain simple linear dynamics, the three-dimensional linear inverted pendulum mode (3D-LIPM). Using a real-time control method based on 3D-LIPM, the Meltran V robot successfully demonstrated 3D dynamic walking without the use of any prepared trajectories.

We are studying a wheeled robot with the ability of going up and down stairs. We previously developed a wheeled robot called a 'variable structure type four-wheeled robot' capable of using over a single step, but it cannot negotiate the stair whose tread is short compared with the wheelbase of the robot. In this paper, we propose a 'biped type leg-wheeled robot' which has ability to pass over stairs with short treads. We also propose a trajectory planning method which uses a nominal static walking trajectory as a reference for the dynamic control of the robot and investigate a trajectory control method using computer simulation. We finally confirm the effectiveness of our proposed method with a successful experiment of going up stairs by this robot.

We develop a wheeled robot, called a 'variable structure type four-wheeled robot' which, with a simple structure, has the ability to pass over a step. We have realized its dynamics for passing over a step of known height. In this paper, we consider the control of dynamic passing over steps of known heights. We propose the method for planning a dynamic trajectory by evaluating the robot's stability against falling down in landing. We also confirm the possibility of detecting the height of the step using an existing rotary encoder without adding any external sensors. As the result, we have successfully realized an experiment of this robot going up steps whose height is unknown within only four seconds.

Cooperative transportation of an object by two or more mobile robots requires each robot to exert an appropriate force to support and move the object, to move along the object, and to maintain its attitude. A mechanically unstable robot, such as a wheeled inverted pendulum, needs to control the force and follow the object as standing stably against the force in an integrated manner. We call this a cooperative behavior. To realize this behavior of a wheeled inverted pendulum, we built control systems to estimate the external force and maintain standing and to exert a specified force. We applied them to a wheeled inverted pendulum. Experiments were conducted of cooperative transportation between a human and the robot. In the future, we aim to realize cooperative transportation by two wheeled inverted pendulums using the control systems.

We have developed a robot called a ' structure type four-wheeled robot' which, with a simple structure, has ability to pass over a step. This robot has only one degree of freedom to change the body's structure and can run on four wheels (4-wheeled mode) or on two wheels like a wheeled inverted pendulum (2-wheeled mode). For the realization of stepping up and down, there are three difficult problems to be solved: stable control in 2-wheeled state, transfer from 4- to 2-wheeled mode (swing-up), and transfer from 2- to 4-wheeled mode (landing). The control method we have developed for a wheeled inverted pendulum is used to control the 2-wheeled state using the signal of a rate gyroscope. In this paper, we mainly discuss control methods for transfers from 4- to 2-wheeled mode and from 2- to 4-wheeled mode. Especially in the transfer from 4- to 2-wheeled state, we make effective use of the driving torque of the motor which change the body's structure. In the transfer from 2- to 4-wheeled state, we have investigated a control method for soft landing to minimize the touch-down impact of the robot which is nonholonomic system. By combining our control methods, we have realized an experiment of passing over a step successfully.