Research Interests

We address issues on skill learning and motor control in humans and robots, and develop learning and control frameworks to achieve complex movements and skills in robotic systems. Our goal is to understand and formulate the principles of motion generation and control design. Our research interest is in the fields of dynamical systems theory, adaptive and optimal control and statistical learning, and their application to robotics.

Optimal Variable Stiffness Control

We address the optimal control problems of robotic systems with variable stiffness actuation including switching dynamics and discontinuous state transitions. Our focus is to consider dynamic tasks that have multiple phases of movement, contacts and impacts with a requirement of exploiting passive dynamics. We develop a systematic methodology to simultaneously optimize control commands, time-varying stiffness profiles and temporal aspect of the movement to exploit the benefits of variable stiffness optimization. Applications to achieving dynamic tasks such as robot brachiation and hopping are explored.

  • J. Nakanishi, A. Radulescu, D. J. Braun and S. Vijayakumar, Spatio-temporal stiffness optimization with switching dynamics, Autonomous Robots (2016) [DOI, Open access]
  • J. Nakanishi, A. Radulescu, and S. Vijayakumar, Spatio-temporal Optimization for Multi-phase Movements: Dealing with Contacts and Switching Dynamics, IEEE/RSJ International Conference of Intelligent Robots and Systems, pp. 5100-5107 (2013) [DOI]
  • J. Nakanishi and S. Vijayakumar, Exploiting Passive Dynamics with Variable Stiffness Actuation in Robot Brachiation, Robotics: Science and Systems, Sydney, Australia, July 9-13 (2012) [Full text at RSS online Proceedings]
variable stiffness control

Operational Space Control

Motivated by the need to bring dexterous and compliant control to complex and highly redundant robots such as humanoid robots, we conduct theoretical and empirical investigations of operational space control algorithms with redundancy resolution. The goal of this study is to examine the practical suitability of operational space control methods for complex high DOF robots. We consider velocity-based, acceleration-based and force-based controllers in the literature, and also develop several new algorithms. We address the practical properties and performance of different approaches, particularly in light of inevitable modeling errors of the robot dynamics.

  • J. Nakanishi, R. Cory, M. Mistry, J. Peters, and S. Schaal, Operational Space Control: A Theoretical and Empirical Comparison, International Journal of Robotics Research, vol. 27, no. 6, pp. 737-757 (2008) [DOI]
operational space control

Internal Model Learning with Adaptive Control

With the requirement of the system's model for motor control in robots and humans performing dynamic motion, we develop a provably stable learning adaptive control framework with statistical learning. Our algorithm employs nonlinear function approximation with automatic structure adaptation and achieves rapid and stable learning. Furthermore, from a perspective of human motor control in computational neuroscience, we address formal stability analysis of the feedback error learning scheme. We derive stability conditions for feedback error learning from the viewpoint of adaptive control.

  • J. Nakanishi, J. Farrell, and S. Schaal, Composite Adaptive Control with Locally Weighted Statistical Learning, Neural Networks, vol. 18, issue 1, pp. 71-90 (2005) [DOI]
  • J. Nakanishi, and S. Schaal, Feedback Error Learning and Nonlinear Adaptive Control, Neural Networks, vol. 17, issue 10, pp. 1453-1465 (2004) [DOI]
internal model learning

Learning and Adaptation of Biped Locomotion

In this study, we develop a framework for learning biped locomotion using dynamical movement primitives based on non-linear oscillators. We use dynamical movement primitives as a central pattern generator (CPG) of a biped robot. Demonstrated trajectories are learned through movement primitives by locally weighted regression, and the frequency of the learned trajectories is adjusted automatically by a frequency adaptation algorithm based on phase resetting and entrainment of coupled oscillators. The role of phase resetting is evaluated for robust walking against external perturbations and environmental changes.

  • J. Nakanishi, J. Morimoto, G. Endo, G. Cheng, S. Schaal, and M. Kawato, Learning from Demonstration and Adaptation of Biped Locomotion, Robotics and Autonomous Systems, vol. 47, no. 2-3, pp. 79-91 (2004) [DOI]
biped robot

Dynamic Movement Primitives for Imitation Learning

The goal of this study is to develop an encoding scheme for complex movements with dynamical systems for imitation learning. Trajectories are represented in a set of nonlinear differential equations with well-defined attractor dynamics equipped with a nonlinear function approximator as a forcing term. Demonstrated movements are learned through locally weighted regression allowing rapid learning. Our approach has desirable properties such as ease of scaling of learned trajectories and online modulation with perceptual coupling and also has been applied to achieving a wide variety of tasks.

  • A. J. Ijspeert, J. Nakanishi, H. Hoffmann, P. Pastor, and S. Schaal, Dynamical Movement Primitives: Learning Nonlinear Attractor Models for Motor Behaviors, Neural Computation, no. 25, issue 2, pp. 328-373 (2013) [DOI]
movement primitives

Task Encoding with Dynamical Systems: Robot Brachiation

In this study, we investigate control strategies to achieve dynamically dexterous behavior of robotic systems with a focus on underactuation. We consider brachiation as an example of interesting form of dynamic locomotion with the help of gravity. The task is encoded as an output of a target dynamical system motivated by the physical insight into the task. We address the problems of swing locomotion, swing-up and forward velocity control of brachiation. In addition, we explore leaping maneuver including flight phase during locomotion.

  • J. Nakanishi, T. Fukuda, and D. E. Koditschek, A Brachiating Robot Controller, IEEE Transactions on Robotics and Automation, vol. 16, no. 2, pp. 109-123 (2000) [DOI]