skip to main content

Title: Adapting control policies from simulation to reality using a pairwise loss
This paper proposes an approach to domain transfer based on a pairwise loss function that helps transfer control policies learned in simulation onto a real robot. We explore the idea in the context of a “category level” manipulation task where a control policy is learned that enables a robot to perform a mating task involving novel objects. We explore the case where depth images are used as the main form of sensor input. Our experimental results demonstrate that proposed method consistently outperforms baseline methods that train only in simulation or that combine real and simulated data in a naive way
Authors:
; ;
Award ID(s):
1724191 1724257 1750649 1763878
Publication Date:
NSF-PAR ID:
10073538
Journal Name:
International Symposium on Experimental Robotics
Sponsoring Org:
National Science Foundation
More Like this
  1. ABSTRACT Introduction Short response time is critical for future military medical operations in austere settings or remote areas. Such effective patient care at the point of injury can greatly benefit from the integration of semi-autonomous robotic systems. To achieve autonomy, robots would require massive libraries of maneuvers collected with the goal of training machine learning algorithms. Although this is attainable in controlled settings, obtaining surgical data in austere settings can be difficult. Hence, in this article, we present the Dexterous Surgical Skill (DESK) database for knowledge transfer between robots. The peg transfer task was selected as it is one of the six main tasks of laparoscopic training. In addition, we provide a machine learning framework to evaluate novel transfer learning methodologies on this database. Methods A set of surgical gestures was collected for a peg transfer task, composed of seven atomic maneuvers referred to as surgemes. The collected Dexterous Surgical Skill dataset comprises a set of surgical robotic skills using the four robotic platforms: Taurus II, simulated Taurus II, YuMi, and the da Vinci Research Kit. Then, we explored two different learning scenarios: no-transfer and domain-transfer. In the no-transfer scenario, the training and testing data were obtained from the samemore »domain; whereas in the domain-transfer scenario, the training data are a blend of simulated and real robot data, which are tested on a real robot. Results Using simulation data to train the learning algorithms enhances the performance on the real robot where limited or no real data are available. The transfer model showed an accuracy of 81% for the YuMi robot when the ratio of real-tosimulated data were 22% to 78%. For the Taurus II and the da Vinci, the model showed an accuracy of 97.5% and 93%, respectively, training only with simulation data. Conclusions The results indicate that simulation can be used to augment training data to enhance the performance of learned models in real scenarios. This shows potential for the future use of surgical data from the operating room in deployable surgical robots in remote areas.« less
  2. We present a strategy for simulation-to-real transfer, which builds on recent advances in robot skill decomposition. Rather than focusing on minimizing the simulation–reality gap, we propose a method for increasing the sample efficiency and robustness of existing simulation-to-real approaches which exploits hierarchy and online adaptation. Instead of learning a unique policy for each desired robotic task, we learn a diverse set of skills and their variations, and embed those skill variations in a continuously parameterized space. We then interpolate, search, and plan in this space to find a transferable policy which solves more complex, high-level tasks by combining low-level skills and their variations. In this work, we first characterize the behavior of this learned skill space, by experimenting with several techniques for composing pre-learned latent skills. We then discuss an algorithm which allows our method to perform long-horizon tasks never seen in simulation, by intelligently sequencing short-horizon latent skills. Our algorithm adapts to unseen tasks online by repeatedly choosing new skills from the latent space, using live sensor data and simulation to predict which latent skill will perform best next in the real world. Importantly, our method learns to control a real robot in joint-space to achieve these high-level tasksmore »with little or no on-robot time, despite the fact that the low-level policies may not be perfectly transferable from simulation to real, and that the low-level skills were not trained on any examples of high-level tasks. In addition to our results indicating a lower sample complexity for families of tasks, we believe that our method provides a promising template for combining learning-based methods with proven classical robotics algorithms such as model-predictive control.

    « less
  3. We develop an approach to improve the learning capabilities of robotic systems by combining learned predictive models with experience-based state-action policy mappings. Predictive models provide an understanding of the task and the dynamics, while experience-based (model-free) policy mappings encode favorable actions that override planned actions. We refer to our approach of systematically combining model-based and model-free learning methods as hybrid learning. Our approach efficiently learns motor skills and improves the performance of predictive models and experience-based policies. Moreover, our approach enables policies (both model-based and model-free) to be updated using any off-policy reinforcement learning method. We derive a deterministic method of hybrid learning by optimally switching between learning modalities. We adapt our method to a stochastic variation that relaxes some of the key assumptions in the original derivation. Our deterministic and stochastic variations are tested on a variety of robot control benchmark tasks in simulation as well as a hardware manipulation task. We extend our approach for use with imitation learning methods, where experience is provided through demonstrations, and we test the expanded capability with a real-world pick-and-place task. The results show that our method is capable of improving the performance and sample efficiency of learning motor skills in amore »variety of experimental domains.« less
  4. An option is a short-term skill consisting of a control policy for a specified region of the state space, and a termination condition recognizing leaving that region. In prior work, we proposed an algorithm called Deep Discovery of Options (DDO) to discover options to accelerate reinforcement learning in Atari games. This paper studies an extension to robot imitation learning, called Discovery of Deep Continuous Options (DDCO), where low-level continuous control skills parametrized by deep neural networks are learned from demonstrations. We extend DDO with: (1) a hybrid categorical–continuous distribution model to parametrize high-level policies that can invoke discrete options as well continuous control actions, and (2) a cross-validation method that relaxes DDO’s requirement that users specify the number of options to be discovered. We evaluate DDCO in simulation of a 3-link robot in the vertical plane pushing a block with friction and gravity, and in two physical experiments on the da Vinci surgical robot, needle insertion where a needle is grasped and inserted into a silicone tissue phantom, and needle bin picking where needles and pins are grasped from a pile and categorized into bins. In the 3-link arm simulation, results suggest that DDCO can take 3x fewer demonstrations tomore »achieve the same reward compared to a baseline imitation learning approach. In the needle insertion task, DDCO was successful 8/10 times compared to the next most accurate imitation learning baseline 6/10. In the surgical bin picking task, the learned policy successfully grasps a single object in 66 out of 99 attempted grasps, and in all but one case successfully recovered from failed grasps by retrying a second time.« less
  5. Simulation provides vast benefits for the field of robotics and Human-Robot Interaction (HRI). This study investigates how sensor effects seen in the real domain can be modeled in simulation and what role they play in effective Sim2Real domain transfer for learned perception models. The study considers introducing naive noise approaches such as additive Gaussian and salt and pepper noise as well as data-driven sensor effects models into simulation for representing Microsoft Kinect sensor capabilities and phenomena seen on real world systems. This study quantifies the benefit of multiple approaches to modeling sensor effects in simulation for Sim2Real domain transfer by their object classification improvements in the real domain. User studies are conducted to address hypotheses by training grounded language models in each of the sensor effects modeling cases and evaluated on the robot's interaction capabilities in the real domain. In addition to grounded language performance metrics, user study evaluation includes surveys on the human participant's assessment of the robot's capabilities in the real domain. Results from this pilot study show benefits to modeling sensor noise in simulation for Sim2Real domain transfer. This study also begins to explore the effects that such models have on human-robot interactions.