A Policy Adaptation Method for Implicit Multitask Reinforcement Learning Problems

In dynamic motion generation tasks, including contact and collisions, small changes in policy parameters can lead to extremely different returns. For example, in soccer, the ball can fly in completely different directions with a similar heading motion by slightly changing the hitting position or the force applied to the ball or when the friction of the ball varies. However, it is difficult to imagine that completely different skills are needed for heading a ball in different directions. In this study, we proposed a multitask reinforcement learning algorithm for adapting a policy to implicit changes in goals or environments in a single motion category with different reward functions or physical parameters of the environment. We evaluated the proposed method on the ball heading task using a monopod robot model. The results showed that the proposed method can adapt to implicit changes in the goal positions or the coefficients of restitution of the ball, whereas the standard domain randomization approach cannot cope with different task settings.

Learning to Shape by Grinding: Cutting-surface-aware Model-based Reinforcement Learning

Object shaping by grinding is a crucial industrial process in which a rotating grinding belt removes material. Object-shape transition models are essential to achieving automation by robots; however, learning such a complex model that depends on process conditions is challenging because it requires a significant amount of data, and the irreversible nature of the removal process makes data collection expensive. This paper proposes a cutting-surface-aware Model-Based Reinforcement Learning (MBRL) method for robotic grinding. Our method employs a cutting-surface-aware model as the object's shape transition model, which in turn is composed of a geometric cutting model and a cutting-surface-deviation model, based on the assumption that the robot action can specify the cutting surface made by the tool. Furthermore, according to the grinding resistance theory, the cutting-surface-deviation model does not require raw shape information, making the model's dimensions smaller and easier to learn than a naive shape transition model directly mapping the shapes. Through evaluation and comparison by simulation and real robot experiments, we confirm that our MBRL method can achieve high data efficiency for learning object shaping by grinding and also provide generalization capability for initial and target shapes that differ from the training data.

Randomized-to-Canonical Model Predictive Control for Real-world Visual Robotic Manipulation

Many works have recently explored Sim-to-real transferable visual model predictive control (MPC). However, such works are limited to one-shot transfer, where real-world data must be collected once to perform the sim-to-real transfer, which remains a significant human effort in transferring the models learned in simulations to new domains in the real world. To alleviate this problem, we first propose a novel model-learning framework called Kalman Randomized-to-Canonical Model (KRC-model). This framework is capable of extracting task-relevant intrinsic features and their dynamics from randomized images. We then propose Kalman Randomized-to-Canonical Model Predictive Control (KRC-MPC) as a zero-shot sim-to-real transferable visual MPC using KRC-model. The effectiveness of our method is evaluated through a valve rotation task by a robot hand in both simulation and the real world, and a block mating task in simulation. The experimental results show that KRC-MPC can be applied to various real domains and tasks in a zero-shot manner.

Phase Portraits as Movement Primitives for Fast Humanoid Robot Control

Currently, usual approaches for fast robot control are largely reliant on solving online optimal control problems. Such methods are known to be computationally intensive and sensitive to model accuracy. On the other hand, animals plan complex motor actions not only fast but seemingly with little effort even on unseen tasks. This natural sense of time and coordination motivates us to approach robot control from a motor skill learning perspective to design fast and computationally light controllers that can be learned autonomously by the robot under mild modeling assumptions. This article introduces Phase Portrait Movement Primitives (PPMP), a primitive that predicts dynamics on a low dimensional phase space which in turn is used to govern the high dimensional kinematics of the task. The stark difference with other primitive formulations is a built-in mechanism for phase prediction in the form of coupled oscillators that replaces model-based state estimators such as Kalman filters. The policy is trained by optimizing the parameters of the oscillators whose output is connected to a kinematic distribution in the form of a phase portrait. The drastic reduction in dimensionality allows us to efficiently train and execute PPMPs on a real human-sized, dual-arm humanoid upper body on a task involving 20 degrees-of-freedom. We demonstrate PPMPs in interactions requiring fast reactions times while generating anticipative pose adaptation in both discrete and cyclic tasks.

An Optimal Assistive Control Strategy based on User's Motor Goal Estimation

In this study, we propose an optimal assistive control strategy that uses estimated user's movement intention as the terminal cost function. We estimate the movement intention by observing human user's joint angle, angluar velocity, and muscle activities for very short period of time. A task-related low-dimensional feature space is extracted from the observed user's movement data. We assume that discrete number of optimal control laws associated to different target tasks are pre-computed. Then, the optimal assistive policy is derived by blending the pre-computed optimal control laws based on the linear Bellman combination method. Coefficients that determine how to blend the control laws are derived based on the low-dimensional feature value that represents the user's movement intention. To validate our proposed method, we conducted basketball throwing tasks. In these experiments, subjects were asked to throw a basketball into a hoop placed at different throwing distances. The distances from the throwing point to the hoop were estimated as the user's movement intention and the optimal control policies were derived by using our proposed method. The results showed that the basketball throwing performances of the subjects were mostly improved.

Efficient Reuse of Previous Experiences to Improve Policies in Real Environment

In this study, we show that a movement policy can be improved efficiently using the previous experiences of a real robot. Reinforcement Learning (RL) is becoming a popular approach to acquire a nonlinear optimal policy through trial and error. However, it is considered very difficult to apply RL to real robot control since it usually requires many learning trials. Such trials cannot be executed in real environments because unrealistic time is necessary and the real system's durability is limited. Therefore, in this study, instead of executing many learning trials, we propose to use a recently developed RL algorithm, importance-weighted PGPE, by which the robot can efficiently reuse previously sampled data to improve it's policy parameters. We apply importance-weighted PGPE to CB-i, our real humanoid robot, and show that it can learn a target reaching movement and a cart-pole swing up movement in a real environment without using any prior knowledge of the task or any carefully designed initial trajectory.

Model-Based Policy Gradients with Parameter-Based Exploration by Least-Squares Conditional Density Estimation

The goal of reinforcement learning (RL) is to let an agent learn an optimal control policy in an unknown environment so that future expected rewards are maximized. The model-free RL approach directly learns the policy based on data samples. Although using many samples tends to improve the accuracy of policy learning, collecting a large number of samples is often expensive in practice. On the other hand, the model-based RL approach first estimates the transition model of the environment and then learns the policy based on the estimated transition model. Thus, if the transition model is accurately learned from a small amount of data, the model-based approach can perform better than the model-free approach. In this paper, we propose a novel model-based RL method by combining a recently proposed model-free policy search method called policy gradients with parameter-based exploration and the state-of-the-art transition model estimator called least-squares conditional density estimation. Through experiments, we demonstrate the practical usefulness of the proposed method.

Efficient Sample Reuse in Policy Gradients with Parameter-based Exploration

The policy gradient approach is a flexible and powerful reinforcement learning method particularly for problems with continuous actions such as robot control. A common challenge in this scenario is how to reduce the variance of policy gradient estimates for reliable policy updates. In this paper, we combine the following three ideas and give a highly effective policy gradient method: (a) the policy gradients with parameter based exploration, which is a recently proposed policy search method with low variance of gradient estimates, (b) an importance sampling technique, which allows us to reuse previously gathered data in a consistent way, and (c) an optimal baseline, which minimizes the variance of gradient estimates with their unbiasedness being maintained. For the proposed method, we give theoretical analysis of the variance of gradient estimates and show its usefulness through extensive experiments.