Pedestrians darting out from blind spots in driver vision are typical scenarios in urban street environments, and conventional autonomous emergency braking systems reach safety limits if sensors do not detect the pedestrian in time to prevent accident or injury. The system must be able to anticipate such potential hazards and to anticipate such pedestrian action. This paper focuses on a pedestrian collision avoidance system that has a “driving-intelligence” model. The model was designed by applying potential field theory using hazard-anticipatory knowledge. The effectiveness of the proposed system is confirmed by computer simulation.

In this paper, we propose a real-time image mosaicing system that uses a high-frame-rate video sequence. Our proposed system can mosaic 512×512 color images captured at 500 fps as a single synthesized panoramic image in real time by stitching the images based on their estimated frame-to-frame changes in displacement and orientation. In the system, feature point extraction is accelerated by implementing a parallel processing circuit module for Harris corner detection, and hundreds of selected feature points in the current frame can be simultaneously corresponded with those in their neighbor ranges in the previous frame, assuming that frame-to-frame image displacement becomes smaller in high-speed vision. The efficacy of our system for improved feature-based real-time image mosaicing at 500 fps was verified by implementing it on a field-programmable gate array (FPGA)-based high-speed vision platform and conducting several experiments: (1) capturing an indoor scene using a camera mounted on a fast-moving two-degrees-of-freedom active vision, (2) capturing an outdoor scene using a hand-held camera that was rapidly moved in a periodic fashion by hand.

This paper deals with an optimal landing motion of a four-link legged robot that minimizes the impact force at the contact point and the joint torques necessary during the motion. The cost function for optimization is given as the weighted sum of the impact force and the joint torques. The configuration of the robot that is close to a singular configuration is advantageous in minimizing the joint torques for a heavy torso, while the configuration where the leg is bent is advantageous in reducing the impact force. This is shown by numerical optimization results with different weights for the cost function and a theoretical analysis of a simplified model of the robot.

Diesel engines effectively reduce automotive CO₂ emissions. Technologies for reducing NO_x emissions are also essential, however, for diesel engine use. The research discussed in this paper focuses on urea-SCR as after-treatment system for converting NO_x, and proposes the urea-SCR control based on NH₃ sensor output. The NH₃ sensor effectively reduces both NO_x, and NH₃ slip. The control adjusts the amount of urea dosed to the catalyst based on changes in NH₃ slip timing and NH₃ concentration in order to compensate for part-to-part variations in the SCR catalyst and catalyst aging. The proposed urea-SCR system using the NH₃ sensor and applied to a test vehicle increased NO_x conversion performance and improved emission robustness over that of conventional control using a NO_x sensor.

In this paper, a robust nonlinear perfect tracking control for a robot arm with uncertainties is proposed by using operator-based robust right coprime factorization approach. In general, there exist unknown modelling errors in measuring structural parameters of the robot arm and external disturbances in real situations. In the present control system design, the effect of the modelling errors and disturbances on the system performance is considered to be uncertainties of the robot arm dynamics. Considering the uncertainties, a robust nonlinear perfect tracking control using operator-based robust right coprime factorization is investigated. That is, first, considering the unknown uncertain plant generates limitations in obtaining the so-called universal stability and tracking conditions, the effect of uncertain plant is compensated by designed operator-based feedback control scheme. Second, a new perfect tracking condition is proposed for improving the trajectory of the robot arm. Finally, the effectiveness of the designed system is confirmed by simulation results.

In conventional reinforcement learning, a reward function influences the learning results, and therefore, the reward function is very important. To design this function considering a task, knowledge of reinforcement learning is required. In addition to this, a reward function must be designed for each task. These requirements make the design of a reward function unfeasible. We focus on this problem and aim at realizing a method to generate a reward without the design of a special reward function. In this paper, we propose a universal evaluation for sensor inputs, which is independent of a task and is modeled on the basis of the indicator of pleasure and pain in biological organisms. This evaluation estimates the trend of sensor inputs based on the ease of input prediction. Instead of the design of a reward function, our approach assists a human being in learning how to interact with an agent and teaching it his/her demand. We recruited a research participant and attempted to solve the path planning problem. The results show that a participant can teach an agent his/her demand by interacting with the agent and the agent can generate an adaptive route by interacting with the participant and the environment.

To expand the range of activities by unmanned aerial vehicles (UAVs), whose use at disaster sites and other dangerous locations has raised high expectations, UAVs must be capable of several functions, e.g., flight both indoor and outdoor environments. This requires making the airframe and major sensors as compact and light-weight as possible. To do so, we are developing uniaxial laser rangefinders to replace conventional scanning laser rangefinders. One developmental approach to uniaxial laser rangefinders and results regarding their validity are reported in this paper. Using a simulated model – a modified scanning laser rangefinder – we discuss the effectiveness of our approach. Specifically, this is to reduce the range drift due to heat, for which we propose using a multiecho to measure two echoes from a reference plate whose distance is known and that is placed outside of the sensor, and the target object, instead using the internal reference plate used in conventional two-dimensional scanning laser rangefinders. We also report the results of experiments verifying its validity.

We propose a motion planning method for accelerating path-tracking task with constant hand speed using redundant manipulator. Tracking-speed should be increased to satisfy the joint torque and joint velocity limits. We propose a method using the state space of redundant manipulators to analyze these limits and to operate motion planning. This state space has a position on a trajectory, redundant pose, and redundant velocity axes. These limits are projected on state space. We consider, motion planning is path-finding problem on state space at a constant hand speed. To plan a faster motion, a constant hand speed is examined by finding a path with limits in state space. We use computer simulation to test and evaluate our proposed method on 3-link planar manipulator. Results demonstrated motion faster than a motion minimizing joint acceleration norm and a motion minimizing driving torque norm. We also verified a problem when our method was applied to an actual manipulator.

In this paper, operator-based sliding-mode nonlinear control is presented for a process with input constraint. After the mathematical model is discussed for the water level process, it is seen that there exist uncertainty and input constraint in the process. Operator-based robust nonlinear control is then considered by using robust right coprime factorization method to eliminate the effect of uncertainty. Furthermore, operator-based sliding-mode nonlinear control is proposed to deal with the constrained input. Finally, simulation and experimental results are given to show the effectiveness of the proposed method.

This paper describes our development of service robots evaluated in real restaurants. Our objective is to develop two types of robot for 1) taking orders from customers and 2) delivering an order to the table. This project is done in collaboration between Bangkok University and MK Restaurants Group Public Company Limited. To achieve this objective, we have applied simple technologies from our experiences at the robot competition called Robocon. We have evaluated each of our robots for six months in a real environment where five branches of the MK restaurant chain are located in the Bangkok area of Thailand, from 2009 to 2012. In the evaluation, robots provided 14,280 services and attracted the interest of 235,680 customers. “Lessons Learned” from this four-year project have been summarized, and should prove useful to similar service robot development projects.

A light weight robotic arm (Jaco) has been interfaced with a novel gesture detection sensor (Leap Motion Controller), substituting complicated conventional input devices, i.e., joysticks and pads. Due to the enhanced precision and high throughput capabilities of the Leap Motion Controller, the unobtrusive measurement of physiological tremor can be extracted. An algorithm was developed to constantly detect and indicate potential user hand tremor patterns in real-time. Additionally a calibration algorithm was developed to allow an optimum mapping between the user hand movement, tracked by the Leap Motion Controller, and the Jaco arm, by filtering unwanted oscillations, allowing for a more natural human-computer interaction.

GEWOS (Healthy Living in an Assisted Environment) is a BMBF-funded R&D-project (budget: 3.6 million €) in which a mechatronic chair for personal health and fitness assistance was developed and evaluated. Besides the chairs of the authors, several industrial partners joined the project team. The mechatronic chair is able to monitor a variety of the user’s/patient’s training activities and vital signs and assists with providing a customized exercise program.

The research project LISA (Living Independently in Südtirol Alto-Adige) investigated the possibilities of embedding assistive functions, systems, and services into wall “terminal” components that enable and support autonomy and independence with respect to Activities of Daily Living (ADLs), and which generate structured environments called Robotic micro-Rooms (RmRs).

Within the ongoing research project, the Chair of Building Realization and Robotics is identifying, developing and validating robotic-based concepts for intra-house mobility, logistics and transfer to support elderly people within their home environment. The project is conducted jointly with several research and industry partners.

In the project USA² (Ubiquitäres und Selbstständiges Arbeiten in einer alternden Gesellschaft), a robotic, mini-factory-like workspace was developed which integrates novel technologies from the field of telepresence, cooperative robotics, seamless interaction, 3D printing and cloud manufacturing.

In anticipation of the aging of society, support equipment is taking an important role in promoting the independence of elderly and disabled citizens, and welfare and other assist devices have been developed to enhance their quality of life (QOL). This development has proceeded based on trial and error, however, dependent on the experiences and attitudes of designer engineers rather than on users. This is because of a lack of information, e.g., on the requirements of the elderly and indices for better evaluating their physical needs. Few such devices currently incorporate operational mechanisms suitable to the physical functions of the elderly and disabled or that consider using their remaining capabilities. An important element in developing support equipment is the concept of power assist – letting users exploit their remaining capabilities while using mechanical means to supplement abilities they may lack. The effects of different kinds of support also are not precisely known. To clarify this, the author developed a link arm to support lacking functions based on data on the movements of unimpaired persons.