Abstract
Autonomous robot learning in unstructured environments often faces the problem that the dimensionality of the search space is too large for practical applications. Dimensionality reduction techniques have been developed to address this problem and describe motor skills in low-dimensional latent spaces. Most of these techniques require the availability of a sufficiently large database of example task executions to compute the latent space. However, the generation of many example task executions on a real robot is tedious, and prone to errors and equipment failures. The main result of this paper is a new approach for efficient database gathering by performing a small number of task executions with a real robot and applying statistical generalization, e.g., Gaussian process regression, to generate more data. We have shown in our experiments that the data generated this way can be used for dimensionality reduction with autoencoder neural networks. The resulting latent spaces can be exploited to implement robot learning more efficiently. The proposed approach has been evaluated on the problem of robotic throwing at a target. Simulation and real-world results with a humanoid robot TALOS are provided. They confirm the effectiveness of generalization-based database acquisition and the efficiency of learning in a low-dimensional latent space.
Highlights
Robot learning, a process where the robot improves its performance by executing the desired task many times to update the principal skill representation, is one of the main technological enablers that can take robots into unstructured environments [1,2]
The consequence is that we reduce the amount of real-world robot executions to acquire a dataset suitable for autoencoder training and learn faster because Reinforcement Learning (RL) is performed in a smaller search space defined by the latent space of the autoencoder
We obtained M results describing the error of each autoencoder network: MSEpos,m, MSEφ,m, MSEθ,m, m = 1, . . . , M, all calculated according to Equations (14), (15), and (16), respectively
Summary
A process where the robot improves its performance by executing the desired task many times to update the principal skill representation, is one of the main technological enablers that can take robots into unstructured environments [1,2]. Robot learning can be a complicated and lengthy process, which requires numerous iterations, trials, and repetitions, all of which might not be safe for the robot or its immediate environment. This is the case when the robot needs to learn a new task from scratch—the search space is too large [3]. This is the case for monolithic problems where only one type of solution is possible (in our practical example, only one way of throwing) [4]. In an action of reaching for an object, one cannot generalize between reaching for an object from the left and from the right, but only separately between examples of reaching from the same class (same side) of reaching movements [11]
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
