There are several ways to be a member of JRL.

Permanent Positions

You can find job opportunities of AIST here. We usually have two rounds of recruitment every year in Spring and Fall. Since the JRL belongs to the department of Information Technology and Human factors (ITH), please check recruitment information of the department. It is recommended to contact before submitting an application.

Visiting Researchers

If you want to stay at JRL as a visiting researcher, please contact us.
JSPS fellows can be hosted at JRL. If you are planning to apply to JSPS Postdoctoral Fellowships for Research in Japan or Invitational Fellowships for Research in Japan, please contact us.

Contractual Positions

We sometimes post job opportunities for contract employees (postdoc and engineer) to here (only in Japanese).

Students/Internships

(Click here for HumanoidLab page)

Students can be hosted at JRL as technical trainees and excellent master/PhD students can be hired as a research assistant of AIST. Although students from any universities can do their internships at JRL, students from the following universities can do their thesis at JRL under supervision of researchers of JRL.
  • University of Tsukuba:
    Since some JRL members are adjunct professors of University of Tsukuba, master and PhD students can do their thesis by enrolling at University of Tsukuba and doing research work at JRL. Humanoid Laboratory is a laboratory of University of Tsukuba hosted by JRL using Cooperative Graduate School System.
  • University of Montpellier:
    PhD students of University of Montpellier can do their thesis at JRL. The JRL is the only one laboratory outside of France where PhD students of University of Montpellier can do their PhD thesis.
  • Other French universities:
    Students from some other French universities can also do a part of their Ph.D. thesis at JRL. Please don't hesitate to contact us for more details.

Open topics

Currently, we are offering the following topics. If you want to do your internship or thesis at JRL, please contact us.
Title Abstract Grade Expected skills
Adaptive torque control Usually the control of robots rely on precise models, however, these can be unavailable or drift over time. Adaptive control techniques will be expected to correct for these modeling errors and compensate for them during the motion. The student will be working on our humanoid robots. master (possibility to extend to PhD)

Preferred: C++, control theory, multibody dynamics

Required: basic calculus, spoken English language

Humanoid robots force control for multi-contact locomotion Locomotion control of legged robots requires force control that relies usually on contact force sensors, however humanoid or legged robot may use parts of their body that are not equipped with these sensors such as elbows and knees. In this work the student will develop and test a control of this kind of contact. Master (possibility to extend to PhD)

Preferred: C++, ROS, control theory, multibody mechanics.

Required: basic geometry and calculus, spoken English language

Safe training of neural networks on real robotic platforms In most cases the training of neural networks to control the motion of humanoid and legged robots, happen in numerical simulation and then transferred to the real robot. In this work the student will design a framework to train neural networks directly on the real robot. They need to ensure the safety and the autonomy of the learning pipeline. Master/PhD

Preferred: C++, Deep learning, ROS, multibody mechanics.

Required: spoken English language

Direct Visual Servoing from farther and faster
  • Robot motion control from camera image brightness
  • Accuracy < 1 mm but 3D trajectories are not ideal
  • Goals:
    • new Hessian construction algorithm
    • hybrid camera field-of-view desired / current images
  • Related projects: DVS-straight, M2SV
Master/PhD C++ programming, mathematical background
Contact-Ready Robust Hydraulic Actuator for Mobile Robots To vastly increase the use of robots, tasks with contacts are inevitable. In this research, we aim to develop robust and force sensitive hydraulic actuators to realize robust and dexterous manipulation on mobile robots. Master Mechanical design, 3D CAD, Basic fluid dynamics, Basic mechatronics
Optimal structural design and structural force sensing for robot systems Force sensitivity is an important feature for robots that co-exist with humans. However, use of torque sensor was limited from increase in weight and size, and lowering of resonance frequency. In this research, we study force sensing technique using structural materials and optimization of sensitivity and structural robustness. Master Mechanical design, Basic electronics, Basic numerical optimization
AI-aided teleoperation to achieve seamless telepresence for non-experts
  • Development of semi-automatic grasping and user intention recognition technologies
  • Efficient combination of machine learning and human guidance to achieve AI-aided teleoperation
  • Related project: SIGRoLe
Master/PhD C++, Python, Linux, machine learning
Pose estimation and mesh reconstruction of objects in the wild
  • Identify objects of interest in the visible field of view without prior knowledge
  • Identify major affordances and display them as AR visual aids
  • Related project: SIGRoLe
Master C++, Python, Linux, computer vision for object pose detection
Enhanced visual feedback for immersive humanod robot teleoperation
  • Achieve an immersive experience for the operator of an avatar in a remote location
  • Use of SLAM and AR to complement visual information outside of the field of view of the robot
  • Related project: SIGRoLe
Master Unity/Unreal, computer vision, SLAM
Navigation and control of a mobile robot with mounted arm
  • Develop autonomous/teleoperated navigation for a mobile robot
  • Develop controller for robotic arm mounted on the mobile robot to perform real-world tasks
  • Related project: ICPS
Master C++, Python, ROS2, Linux, planning and control of mobile robots and manipulators
Impact management for humanoid robots using nonsmooth controllers
  • Develop and implement non-smooth controllers for dealing with impacts
  • Realize hammering, nail-gun and jumping tasks using humanoid robots
  • Related project: ImpAdm
Master/PhD C++, Linux, dynamics, non-linear control theory
Simulation and control of an inflatable module mockup
  • Simulate the dynamics of a cylindric inflatable module
  • Control the inflation sequence of the chambers of the module to achieve controlled rolling
  • Validate control using a small mockup
Master C++, Linux, CAD design, FEM-based simulation, fluid mechanics, non-linear control theory
Responsive multi-contact locomotion
  • Passing through narrow space using not only legs but also arms freely
  • Efficient search for contact transition and whole-body motion
  • Efficient solver for non-linear model predictive control
Master/PhD Dynamics, Control theory, Optimization, Graph search, Programming(c++), (Machine learning)
Data-driven quadruped controller toward outdoor exploration
  • Data-driven controller for quadruped robot to traverse uneven outdoor environments
  • Multimodal sensor fusion robust to environmental changes
  • SimToReal that requires no parameter fine-tuning
Master Reinforcement learning, RNN, DNN, Programming(python, c++)
Construction and Retrieval of Multi-contact Motion Library for Humanoid Robot We offer research about the methodology of construction and retrieval of motion library to achieve on-site multi-contact motion planning for a humanoid robot. We want to clarify the structure of the motion library, which can preserve the feasible whole-body motion of a humanoid robot. We also want to develop the online adaptation method to modify preserved motion to satisfy kinematics and dynamics constraints in an unknown environment. PhD Basic knowledge of robotics, Software development by C++ and Python with ROS
Vision-based deformation of human body deformation for safe robot interaction
  • Understanding the impact of a robot physical contact with a human is hard
  • Robot visual perception offers a potential of thoroughness and accuracy for such task
  • Goals:
    • Leverage physics-based modeling to train visual spatio-temporal features
    • Frugal and real-time processing
    • Integrate in human robot physical interaction (e.g. assistive robotics)
Master/PhD mathematical and mechanical background in 3D geometry, optimization, training, C++ / python programming
Human body visual tracking leveraging biomechanical modelling
  • State-of-the-art biomechanical human tracking is marker-based optical motion capture
  • Markerless human visual trackers are more flexible but with puppet-like models
  • Goals:
    • Leverage biomechanical models in markerless human visual tracking
    • Improve human dynamics prediction
    • Integrate in human robot physical interaction (e.g. load carrying)
Master mathematical and mechanical background, 3D geometry, optimization, training, C++ / python programming