Reinforcement learning algorithms have proven to be capable of solving complicated robotics tasks in an end-to-end fashion without any need for hand-crafted features or policies. Especially in the context of robotics, in which the cost of real-world data is usually extremely high, Reinforcement Learning solutions achieving high sample efficiency are needed. In this paper, we propose a framework combining the learning of a low-dimensional state representation, from high-dimensional observations coming from the robot’s raw sensory readings, with the learning of the optimal policy, given the learned state representation. We evaluate our framework in the context of mobile robot navigation in the case of continuous state and action spaces. Moreover, we study the problem of transferring what learned in the simulated virtual environment to the real robot without further retraining using real-world data in the presence of visual and depth distractors, such as lighting changes and moving obstacles. A video of our experiments can be found at: https://youtu.be/rUdGPKr2Wuo.

Reinforcement Learning has been able to solve many complicated robotics tasks without any need for feature engineering in an end-to-end fashion. However, learning the optimal policy directly from the sensory inputs, i.e the observations, often requires processing and storage of a huge amount of data. In the context of robotics, the cost of data from real robotics hardware is usually very high, thus solutions that achieve high sample-efficiency are needed. We propose a method that aims at learning a mapping from the observations into a lower-dimensional state space. This mapping is learned with unsupervised learning using loss functions shaped to incorporate prior knowledge of the environment and the task. Using the samples from the state space, the optimal policy is quickly and efficiently learned. We test the method on several mobile robot navigation tasks in a simulation environment and also on a real robot. A video of our experiments can be found at: https://youtu.be/dgWxmfSv95U.

This paper presents a new free-flight controller for aerial manipulators, unmanned aerial vehicles endowed with a robotic manipulator. The control strategy exploits the dynamics of the manipulator to improve the tracking performance and maneuverability of the UAV by expanding its flight envelop. The controller increases the spectrum of deployable manipulators, which otherwise are limited to be light-weighted manipulators and with a dynamics that do not significantly affect the UAV. The effectiveness and applicability of the proposed controller is verified through simulations and experiments.

This paper presents a versatile control architecture for aerial robots in interactive tasks. The control architecture is characterized by its unique capability of varying the apparent impedance of the controlled aerial robot as well as the interaction force, when in contact. This work finds its way in various applications where different impedance and interaction force controllers provide high task performances as well as safety. The feasibility and effectiveness of the proposed controller are demonstrated by experimental results preformed on a quadrotor aerial robot.

This paper presents a generic hierarchical passive teleoperation control architecture that effectively addresses the issues of workspace incompatibility and precision, as well as other classical and peculiar challenges. More specifically, the control scheme consists of a user-defined variable scale mapping, a variable impedance master controller, and a virtual slave system. The port-based modeling framework has been extensively used in our formulation, providing more insight about energetic flows in the system that are particularly useful for the design of a passive controlled system. Moreover, various practical considerations that are required for the effective usage of the control architecture are discussed. The achieved better precision and overall task performance have been validated and verified by elaborate simulations and experiments.

This paper describes the theory and practice for a stable haptic teleoperation of a flying vehicle. It extends passivity-based control framework for haptic teleoperation of aerial vehicles in the longest intercontinental setting that presents great challenges. The practicality of the control architecture has been shown in maneuvering and obstacle-avoidance tasks over the internet with the presence of significant time-varying delays and packet losses. Experimental results are presented for teleoperation of a slave quadrotor in Australia from a master station in the Netherlands. The results show that the remote operator is able to safely maneuver the flying vehicle through a structure using haptic feedback of the state of the slave and the perceived obstacles.

In this paper, we present haptic teleoperation of underactuated unmanned aerial vehicles by providing a multidimensional generalization of the virtual slave concept. The proposed control architecture is composed of high-level and low-level controllers. The high-level controller commands the vehicle to accomplish specific tasks and renders both the state and the environment of the vehicle to the operator through haptic feedback. The low-level controller interprets the command signals from the operator, regulates the dynamics of the vehicle and feeds back its state to the high-level loop. Passivity of the teleoperation loop is always ensured independently of the choice of implementation of the low-level controller and the configuration of the flying hardware by a passivity-enforcing supervisor, which associates every action of the slave with an energy expense that can only be made available from a multi-state energy tank. The effectiveness of the proposed algorithm is illustrated with simulations and experimental tests.

In this paper, we present a haptic teleoperation control algorithm for unmanned aerial vehicles, applying a kinetic scrolling-based position mapping. The proposed algorithm overcomes the master workspace limitations and enables to teleoperate the aerial vehicle in unbounded workspace in a fast and intuitive manner. Moreover, it provides high precision to teleoperation tasks. Simulation and experimental results validating the applicability and effectiveness of the proposed algorithm are also presented.

This video presents practical realizations and comparison between three different haptic tele-control algorithms of aerial vehicles. These strategies, besides addressing the classical issues of stability and transparency, provide different alternatives for overcoming challenges that are peculiar to haptic teleoperation of aerial vehicles. The experimental results show the performance and effectiveness of the proposed control algorithms even in the presence of significant time delays.

This paper deals with the bilateral teleoperation of underactuated aerial robots by means of a haptic interface. In particular, we propose a switching-based state mapping and control algorithm between a rate-based passive controller, which addresses the workspace incompatibility between the master and slave systems, and a pose-based passive controller, which is required for precise operation. The overall control architecture provides the possibility of changing the scaling factor of the mapping online, while preserving the passivity of the complete system. In our formulation, we use the port-Hamiltonian framework, in which energetic considerations play a determinant role for passivity and, thereby stability of the overall system. Simulation and experimental results illustrating the effectiveness of the proposed algorithm are also presented.

In this paper, we propose a generic model and a controller design for a class of underactuated aerial vehicles, namely for unmanned aerial vehicles whose primary support against gravity is thrust. The approach followed is based on energetic consideration and uses the formalisms of port-Hamiltonian systems and bond graphs. The controller is designed for both stabilization during hovering and for trajectory tracking tasks. The competency of the model and the performance of the controller are validated in simulation.