Vision based behavior verification system of humanoid robot for daily environment tasks

Okada, Kei; Kojima, Mitsuharu; Sagawa, Yuichi; Ichino, Toshiyuki; Satô, Kenji; Inaba, Masayuki

doi:10.1109/ichr.2006.321356

Cited by 85 publications

(47 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are many research efforts aimed at enabling robots to perform household tasks such as cooking and serving food in daily-life environments [1][2][3][4][5]. Reaching target objects is the beginning of many such household tasks.…”

Section: Introductionmentioning

confidence: 99%

Real-time whole-body motion generation using torso posture regression and center of mass

et al. 2018

View full text Add to dashboard Cite

For household humanoid robots, reaching as much workspace as possible with their hands is an important issue because the locations of target objects may range from the floor to above the robot's head. At the same time, to adapt to the constantly-changing household environment, inverse kinematics for the whole body must be solved in real time. In this paper, to achieve real-time motion generation for a humanoid robot, we propose a method of separating the inverse kinematics calculation into simpler problems. Using regression to estimate the torso orientation, we independently solve inverse kinematics for the lower body and both arms. First, using the target pose of both hands as input, we calculate the orientation of the torso and determine the target position of the center of mass considering the reachability of both arms. At each control step, we calculate the joint angles of the lower body from the position of the center of mass, feet poses, and torso orientation. Then, we calculate the joint angles of both arms. In experiments, we apply the proposed method to a human-size humanoid robot for reaching low-height positions while hunkering down. The proposed inverse kinematics solver is ten times faster than the numerical solution using the Jacobian matrix. We also verify the applicability of the proposed method using a sequence of random target positions for the hands as input.

show abstract

Section: Introductionmentioning

confidence: 99%

Real-time whole-body motion generation using torso posture regression and center of mass

et al. 2018

View full text Add to dashboard Cite

show abstract

“…Many researchers around the world address this problem with different approaches such as the behavior based approach [5], the tele-operation approach [6] and the cognitive learning approach [7] and so on. Among them, we have been developing a humanoid system based on knowledge based vision-guided robot system, which is archived through the development of three components: 1) Manipulation knowledge based whole body motion generation system [8], 2) Visual feature knowledge based 3D object recognition system [9], 3) Vision based environment and behavior verification system by using both manipulation and visual feature knowledge [10].…”

Section: Introductionmentioning

confidence: 99%

Task guided attention control and visual verification in tea serving by the daily assistive humanoid HRP2JSK

Okada

Kojima

Tokutsu

et al. 2008

2008 IEEE/RSJ International Conference on Intelligent Robots and Systems

View full text Add to dashboard Cite

Abstract-This paper describes daily assistive task experiments that conducting on the HRP2JSK humanoid robot. We present overall action and recognition integrated system design to realize daily assistive behaviors autonomously and robustly, along with the demonstration that the HRP2JSK pours tea from a bottle to a cup and wash it after human drink it. To obtain autonomy and robustness, visual recognition and behavior control through perception information are important.The significant issue tackled in this paper is what kind of task relevant knowledge is required for daily assistive task humanoids. Reducing search area is well-known technique to increase robustness, however, what kind of information should we embed in the robot is still the open problem, and in the humanoid case, the system has to cope with both manipulation and navigation task.In this paper, we classified prediction based attention control based on following three task relevant knowledge: 1) Predicted search area to restrict potential object location in the recognition process, 2) predicted attention area to restrict imageprocessing area and 3) predicted visual features to eliminate mismatch. Task relevant knowledge is also used for vision guided behavior controls including 1) visual self-localization to recognize the position of a robot, 2) visual object localization to update the object location to generate behaviors and 3) visual behavior verification to confirm the success of the motion, are shown for adapting the planned motion to the real environment.Finally, we demonstrated a tea service task by a humanoid robot. This task was repeated many times as presses or lab tourists demanded. Through this experience, we concluded that the robustness of the developed system reached to a satisfactory level.

show abstract

“…Existing work relating to liquids either uses coarse simulations that are disconnected to real liquid perception and dynamics [3,4] or constrained task spaces that bypass the need to perceive or reason directly about liquids [5,6,7,8,9]. While some of this work has dealt with pouring, none of it has attempted to directly perceive liquids from raw sensory data.…”

Section: Related Workmentioning

confidence: 99%

Towards Learning to Perceive and Reason About Liquids

Schenck

Fox

2017

Springer Proceedings in Advanced Robotics

View full text Add to dashboard Cite

Abstract. Recent advances in AI and robotics have claimed many incredible results with deep learning, yet no work to date has applied deep learning to the problem of liquid perception and reasoning. In this paper, we apply fully-convolutional deep neural networks to the tasks of detecting and tracking liquids. We evaluate three models: a single-frame network, multi-frame network, and a LSTM recurrent network. Our results show that the best liquid detection results are achieved when aggregating data over multiple frames and that the LSTM network outperforms the other two in both tasks. This suggests that LSTM-based neural networks have the potential to be a key component for enabling robots to handle liquids using robust, closed-loop controllers.

show abstract

Vision based behavior verification system of humanoid robot for daily environment tasks

Cited by 85 publications

References 15 publications

Real-time whole-body motion generation using torso posture regression and center of mass

Real-time whole-body motion generation using torso posture regression and center of mass

Task guided attention control and visual verification in tea serving by the daily assistive humanoid HRP2JSK

Towards Learning to Perceive and Reason About Liquids

Contact Info

Product

Resources

About