Shinkichi Inagaki

Modeling and Recognition of Driving Behavior Based on Stochastic Switched ARX Model

This paper presents the development of the modeling and recognition of human driving behavior based on a stochastic switched autoregressive exogenous (SS-ARX) model. First, a parameter estimation algorithm for the SS-ARX model with multiple measured input-output sequences is developed based on the expectation-maximization algorithm. This can be achieved by extending the parameter estimation technique for the conventional hidden Markov model. Second, the developed parameter estimation algorithm is applied to driving data with the focus being on drivers collision avoidance behavior. The driving data were collected using a driving simulator based on the cave automatic virtual environment, which is a stereoscopic immersive virtual reality system. Then, the parameter set for each driver is obtained, and certain driving characteristics are identified from the viewpoint of switched control mechanism. Finally, the performance of the SS-ARX model as a behavior recognizer is examined. The results show that the SS-ARX model holds remarkable potential to function as a behavior recognizer.

Wave CPG model for autonomous decentralized multi-legged robot: Gait generation and walking speed control

Abstract In this paper, we propose a method to control gait generation and walking speed control for an autonomous decentralized multi-legged robot by using a wave Central Pattern Generator (CPG) model. The wave CPG model is a mathematical model of nonlinear oscillators and generates rhythmic movements of the legs. The gait generation and the walking speed control are achieved by controlling the virtual energy of the oscillators (Hamiltonian). A real robot experiment showed the relationship to the Hamiltonian, the actual energy consumption and the walking speed, and the effectiveness of the proposed method was verified.

Nonintrusive appliance load monitoring based on integer programming

電気機器ごとの稼働実態は，電力の需要予測や生活アシストシステムの構築の上で，非常に有用な情報である。その稼働状態を推定するシステムは一般に侵入型と非侵入型のモニタリングシステムに分類される。前者は対象とする電気機器のそれぞれに稼働状態を把握する装置を取り付け，機器ごとの稼働実態を調査する方法である。この方法には，多数の装置を屋内配線の各所に設置するため，装置そのものや設置工事にかかるコストが大きいという問題点がある。これに対して，非侵入型モニタリングシステムとは，建物への給電線の入り口などの配線の上流位置での観測情報から，下流の電気機器の稼働状態を把握するものである。非侵入型モニタリングでは比較的工事のしやすい配線の上流だけに装置を取り付ければよく，装置や設置工事にかかるコストを低く抑えることができる。非侵入型モニタリングシステムに関して，電力のステップ上の変化に注目した方法

CPG model for autonomous decentralized multi-legged robot system—generation and transition of oscillation patterns and dynamics of oscillators

Abstract A central pattern generator (CPG) model is proposed for the gait-pattern generation mechanism of an autonomous decentralized multi-legged robot system. The topological structure of the CPG is represented as a graph on which two time evolution systems, the Hamilton system and a gradient system, are introduced. The CPG model can generate oscillation patterns depending only on the network topology and can bifurcate different oscillation patterns according to the network energy, which means that the robot can generate gait patterns by connecting legs and transit gait patterns according to such parameters as the desired speed.

Modeling and Recognition of Human Driving Behavior based on Stochastic Switched ARX model

This paper presents a development of the modeling of the human driving behavior based on the expression as Stochastic Switched ARX model (SS-ARX) focusing on the driver’s collision avoidance behavior. First, the parameter estimation technique for the SS-ARX model is introduced based on the EM algorithm. This can be achieved by extending the parameter estimation technique for conventional Hidden Markov Model (HMM). Second, the parameter estimation technique is applied to the collected driving data, and find parameter set for each driving data. The driving data are collected by using the three-dimensional driving simulator based on CAVE, which provides stereoscopic immersive vision. Finally, the performance of the SS-ARX model in the case of using as the recognizer is examined. The results show the high potential ability of the SS-ARX model as the behavior recognizer.

Identification of Probability weighted multiple ARX models and its application to behavior analysis

This paper proposes a Probability weighted ARX (PrARX) model wherein the multiple ARX models are composed by the probabilistic weighting functions. As the probabilistic weighting function, a ‘softmax’ function is introduced. Then, the parameter estimation problem for the proposed model is formulated as a single optimization problem. Furthermore, the identified PrARX model can be easily transformed to the corresponding PWARX model with complete partitions between regions. Finally, the proposed model is applied to the modeling of the driving behavior, and the usefulness of the model is verified and discussed.

Analysis and synthesis of driving behavior based on mode segmentation

This paper presents the development of the modeling of human driving behavior based on an expression as a hybrid system (HS) focusing on the driverpsilas vehicle following task. In our modeling, a relationship between the diverpsilas sensory information and the output of driver is expressed by the piecewise ARX (PWARX) model, which is a class of the HS. As the sensory input, the range between vehicles, range rate, and time derivative of the area of the back of the preceding vehicle (called KdB) are considered. Also, the pedal operation is considered as the output. The identification problem for the PWARX model is solved using the clustering based technique. By introducing the PWARX model, it becomes possible to find not only parameters appearing in the operation in each mode but also parameters in the logical switching (decision making) conditions among them from the measured driving data. Furthermore, the obtained model is exploited for the design of assist system especially focusing on the switching condition of the ON/OFF of assist. The usefulness of the proposed assist system is confirmed by experiments.

Follow-the-Contact-Point gait control of centipede-like multi-legged robot to navigate and walk on uneven terrain

This paper proposes a novel locomotion control scheme of centipede-like multi-legged robot, which is called Follow-the-Contact-Point (FCP) gait control. A centipede-like multi-legged robot is composed of segmented trunks which have a pair of legs and are connected with fore and/or rear ones by joints. This control scheme realizes locomotion control of multi-legged robot on uneven terrain with perfectly decentralized manner. The main concept of the control scheme is to relay the contact points from the fore leg to the rear leg. By creating contact points of the first legs adequately on the environment, the robot can climb over obstacles and be navigated successfully. Finally, the result of physical simulation of a 20-legged robot shows the availability of the proposed method.

Design of man-machine cooperative nonholonomic two-wheeled vehicle based on impedance control and time-state control

This paper presents a new control methodology for a nonholonomic electric two-wheeled vehicle wherein the autonomous and man-machine cooperative controls are synthesized. In the proposed control scheme, the ‘autonomous control’ and the ‘man-machine cooperative control’ are designed by synthesizing time-state control and impedance control. The time-state controller tries to reduce the machines deviation from the guideline, the impedance controller, on the other hand, generates power to assist the operators maneuver. Furthermore, experimental results are shown to demonstrate the usefulness of the proposed strategy.

Maximum likelihood estimation of Departure and Travel Time of Individual Vehicle using statistics and dynamic programming

Electric Vehicles (EVs) and Plug-in Hybrid Vehicles (PHVs) generally equip a battery of high capacity. Cars such as EVs and PHVs are expected to work not only as transportation devices, but also as power storages. However, in order to use the battery effectively, we need to know the future Profile of the Departure and Travel Time (PDTT) of the car. This paper presents an estimation method of the PDTT of the car over one day from the present time based on the Statistics of the Departure and Travel Time (SDTT) and dynamic programming. The prediction problem of PDTT of the car is formulated as a maximum-likelihood estimation problem under the condition that the SDTT is available. In order to find a global optimal solution within a reasonable computational cost, first of all, a Markov model representing all possible PDTT of the car is derived from the SDTT. Then, the dynamic programming is applied to find the most likely PDTT of the car. The usefulness of the proposed method is evaluated by numerical experiments, wherein the SDTT is created by real driving data.