Iori Kumagai

Humanoid full-body controller adapting constraints in structured objects through updating task-level reference force

Manipulation of structured objects connected to the environment by a kinematics chain involves two problems: (a) The objects have movable directions and unmovable directions. An undesired reaction force in the unmovable directions prevents a robot from successful manipulation; (b) The reaction forces from the objects could fluctuate during manipulation. Related works have enabled robots to manipulate objects by integrating position control in movable directions and force control in unmovable directions at the hands. However, in the case of a humanoid robot, too large undesired reaction forces in movable directions cause the robots falling down and slipping. In this paper, we propose a controller system controlling reaction forces at the hands and successively updating reference forces based on reaction forces. For problem (a), we apply force control both to the movable and unmovable directions in order to satisfy both maintaining full-body balance and achieving manipulation. For problem (b), the update of the reference forces enables the humanoid robot to adapt to fluctuation of the reaction forces. We show experimental results on the cmanipulating four doors and a drawer.

Development of humanoid robot system for disaster response through team NEDO-JSK's approach to DARPA Robotics Challenge Finals

This paper presents Team NEDO-JSKs approach to the development of novel humanoid platform for disaster response through participation to DARPA Robotics Challenge Finals. This development is a part of the project organized by New Energy and Industrial Technology Development Organization. Technology for this robot is based on the recent research of high-speed and high-torque motor driver with water-cooling system, RTM-ROS inter-operation for intelligent robotics, and generation of full-body fast dancing motion, due to the generic 10 years research of HRP-2 as a platform humanoid robot. Development target is the robot support in a variety of unsafe human tasks teleoperated by humans in case of a disaster response, equipped with body structure capability for use of human devices and tools in human environment, performance for dynamic full-body actions covering human-sized speed and power, and basic function for intelligent and integrated robot platform system for performing various tasks independently. we also describes NEDO-JSK teams approach to design methodology for robot hardware and architecture of software system and user interface for DRC Finals as a test case of disaster response.

Multi-layered real-time controllers for humanoid's manipulation and locomotion tasks with emergency stop

This paper describes a practical method to construct real-time controllers to achieve locomotion and manipulation tasks with a humanoid robot. We propose a method to insert emergency stop functionality to each layer to avoid robots falling down and joint overloads even if recognition and planning error exist. We explain implementation of multi-layered real-time controllers on HRP2 robot and application to several manipulation and locomotion tasks. Finally, we evaluate emergency stop functionality in several manipulation tasks.

Development of a full body multi-axis soft tactile sensor suit for life sized humanoid robot and an algorithm to detect contact states

Recognizing environmental contact on whole body of a humanoid robot can be very advantageous to work with people in humans environment. In the tasks with environmental contacts, it is important as an interface with the environment to detect pushing, shearing and twist on the whole body of a robot such that it gets to know its current state and what to do next. In this paper, we describe a full body soft tactile sensor suit for a humanoid robot and an algorithm to calculate pushing, shearing, and twist for each sensor unit. These sensors are small muti-axis sensors with urethane structure and they can be placed densely on the body of a humanoid robot. We arranged 347 multi-axis soft tactile sensors on a humanoid robot imitating a human tactile sense to detect contact states. Then, we calculate a deformation vector for each muti-axis soft tactile sensor and detect the three contact states using deformation moment and average of deformation vectors in the contact surface consisting of soft tactile sensors. Finally, we confirmed the validity of the full body tactile suit and contact state detector by experiments of sitting on a wheelchair and passing object between a human and a robot.

Achievement of recognition guided teleoperation driving system for humanoid robots with vehicle path estimation

In the wake of the DARPA Robotics Challenge, the task for robots to drive vehicles has been expected to be a method for robots to transport themselves to the disaster site where it is hazardous for humans to approach. In the driving task, it is important for the robot to estimate the path of the vehicle and select an appropriate path for navigation through unknown obstacles, even under limited communication with an operator. It is also necessary for robots to suggest the estimated path of vehicle to an operator to deal with unforeseen circumstances. Therefore, we propose a recognition guided teleoperated driving system for robots to drive vehicles in disaster sites with estimated vehicle path based on steering angle and vehicle model. First, we show model based steering and pedaling strategy to achieve the target steering angle for desired path. Next, we propose a vehicle path estimation and a local planner that can suggest a traveling path according to the surroundings. We integrated them into a teleoperation system for bandwidth limited environments as recognition guidance. Finally, we show the effectiveness of our driving system by conducting field driving experiments with three different robots: JAXON, STARO and HRP-2.

Implementation of a robot-human object handover controller on a compliant underactuated hand using joint position error measurements for grip force and load force estimations

Object handover is a basic task in many human-robot interactive scenarios and therefore, it is important for assistive robots to be able to perform proper handovers. We previously designed a human-inspired grip-force-varying handover controller for a robot giver and showed on a Willow Garage PR2 robot that the controller yields human-like and human-preferred handovers. The PR2 robot had a non-compliant fully-actuated gripper. However, recently, compliant underactuated grippers have been gaining more popularity. Although compliant underactuated grippers can provide more flexibility in manipulation, it is generally difficult to accurately measure and control the amount of applied grip force. In this paper, we present an implementation of the human-inspired handover controller on a Kawada Industries HRP4R robot, which has compliant underactuated hands, using joint position error measurement for estimating the amount of applied grip force. Through an experiment, we show that we are able to achieve safe, smooth, and intuitive robot-human handovers despite the lack of accurate grip force control on our robot.

Whole body joint load reduction control for high-load tasks of humanoid robot through adapting joint torque limitation based on online joint temperature estimation

In the field of assistive and disaster response robotics, robots must perform long term or momentary high-load tasks, such as holding heavy objects or climbing up and descending from a high step in environments with a lot of disturbances. In such cases, the robot joints could potentially break due to an unintended load during high-load tasks in environments where detection of contact forces is difficult. In this paper, we propose a method for reducing the loads on the joints by limiting the joint torque dynamically based on temperature estimation of these joints. Since joint failure is caused by overheating of a motor, it is important to guarantee that a joint motor temperature remains within the safe limits. Our joint load reduction control is essentially a torque limitation method which bases on adapting the maximum torque given the temperature predictions. Such predictions are extracted from the motor thermal model. To do so, we establish a relationship between the joint temperature and the joint torque. The robot uses such relationship to predict the temperature rise as well as the maximum allowable torque. Next, this maximum torque is feeded to a torque controller in order to achieve load reduction. We experimentally tested our method and confirmed that specific high-load tasks can be achieved even in environments where unintended loads occur.

Achievement of localization system for humanoid robots with virtual horizontal scan relative to improved odometry fusing internal sensors and visual information

To achieve tasks in unknown environments with high reliability, highly accurate localization during task execution is necessary for humanoid robots. In this paper, we discuss a localization system which can be applied to a humanoid robot when executing tasks in the real world. During such tasks, humanoid robots typically do not possess a referential to a constant horizontal plane which can in turn be used as part of fast and cost efficient localization methods. We solve this problem by first computing an improved odometry estimate through fusing visual odometry, feedforward commands from gait generator and orientation from inertia sensors. This estimate is used to generate a 3D point cloud from the accumulation of successive laser scans and such point cloud is then properly sliced to create a constant height horizontal virtual scan. Finally, this slice is used as an observation base and fed to a 2D SLAM method. The fusion process uses a velocity error model to achieve greater accuracy, which parameters are measured on the real robot. We evaluate our localization system in a real world task execution experiment using the JAXON robot and show how our system can be used as a practical solution for humanoid robots localization during complex tasks execution processes.

Evaluation-controlling mechanism of perception, planning, and execution for a life-sized humanoid robot

It is necessary for a robot system to control accuracy and region-of-interest according to safety, specified time limit, and available computer resources. If enough duration for a task is provided, robot can use more accurate perception and planners producing higher rate of success. In the research and development of robot systems, these relationship between accuracy and computation time has been tuned carefully by hand. Appropriate sensors, algorithms and parameters are necessary to be chosen by system implementers. In this paper, a constitution method being able to control and evaluate perceptions and actions for robot systems is proposed towards implementing robot systems with automatic parameter controls of perceptions and planners corresponding to specified time limits and success rate.

Creating socially acceptable robots: Leaning grasp configurations for object handovers from demonstrations

Object handover is a basic task that is found in many human-robot cooperation scenarios. If we are to build socially acceptable robots, we need to enable robots to perform handovers properly. In this paper, we discuss some of the social implications of proper robot-human handovers, and we focus on the challenge of determining a proper grasp configuration when handing over an object. We propose a framework that enables a robot to learn proper grasp configurations for handovers through observations. Our aim is to eliminate the need for manually specifying grasp configurations information to the robot, and allow generalization of handover grasp configurations for known objects to unknown objects. We are currently implementing our proposed framework onto an HRP4R robot, and we discuss about our plans for conducting user studies to evaluate our system upon its completion.