In the last period, we continued the work on the real-time 3D reconstruction of the environment. To this end, we integrated and evaluated several stereo depth reconstruction algorithms, namely the "block-matcher" and the "semi-global-matcher" implementation as well as the "efficient large scale stereo" implementation. By rigidly fixing a laser scanner to the stereo sensor, we were able to map the laser rays from the laser frame into the camera frames of the stereo sensor in order to obtain a ground truth for the depth estimation. The next steps are to finish the evaluation of the depth reconstruction algorithms and, in a next step, to combine these depth maps into a 3D map.

During the last period, we improved the bilateral scheme when a 2D laser scanner (coupledwith an IMU) is used to provide information on the environment. We also improved the outer-control scheme of the UAV by exploiting the induced drag in the control design (accepted in IEEE-IROS). We did the last version for the IEEE-Transaction on Mechatronics (accepted as regular paper) and sent the final version of the Chapter "Bilateral Haptic Teleoperation of an Industrial Multirotor UAV" that will appear in Springer Series "STAR".

The current status of the project is that the UAV can be teleoperated in structured environments using the planar 2D laser scanner as the exteroceptive sensing modality. Since the 2D laser scanner only provides a horizontal cross-section of the environment, we make the assumption that the environment consists of vertical walls for localization and obstacle avoidance of the UAV as well as force feedback to the user.

In order to perform flights in an industrial environment (which are usually unstructured three-dimensional environments) we are currently working of adapting our teleoperation scheme by incorporating a visual-inertial stereosensor. This sensor provides both the position of the UAV with respect to the environment as well as a dense depth map of the environment that is observed by the camera.

During the last period, the work on the compensation of aerodynamic drag coefficients, described in the previous report, was continued.  In a first step, the lumped aerodynamic drag coefficient was identified in a Vicon motion tracking system and then tested to show we can improve the transient of the closed-loop system.  
    
Following up on our earlier work on laser-based teleoperation, we now intend to extend this scheme to be employed in unstructured environments. Using the laser-scanner, we made the assumption that the UAV is operated in an environment consisting only of vertical walls. This restricts the use of the UAV to structured environments, e.g. in office environments or while following vertical walls. As outlined in the report "Technical Objectives of TUAV Project and Sensor Evaluation ", we therefore intend to use a visual-inertial stereo-camera system for both localization and obstacle detection of the UAV within unstructured environments. At the industrial project partner Skybotix, such a sensor is available (for more details on the sensor, please refer to the previously mentioned report).  Currently, this sensor is being integrated on the UAV. More precisely, a vibration-damped sensor mount is currently designed for fixation of the sensor of the UAV platform. In parallel, we started to integrate the pre-existing algorithms in our teleoperation framework. Currently, we are evaluating whether a sparse point-cloud (generated from a set of stereo images) provides sufficient information for stable and robust obstacle avoidance.

During the last period, the work was devoted in improving the laser-based wall-following. In particular, the 2D laser range scanner along with attitude information  from the onboard IMU is used to estimate the heading and the distance of the UAV with respect to a wall. This described wall-following scheme, together with the pressure-based altitude control was evaluated in experiment that are disseminated in the current revision of the IEEE Transaction on Mechatronics paper (Hardware and Software Architecture for Nonlinear Control of Multirotor Helicopters) currently under review.

Beginning of the Year 2013, first order aerodynamic drag effects of the UAV were investigated to both improve control- and state estimation algorithms. These first order drag effects originate in the interaction of the rotor blades with the relative airflow. It gathers two terms 'blade flapping' and 'induced drag', both were identified and exploited in a new control scheme to improve the overall behavior of the UAV.

The main work done within the last pariod (October-November) to the extension of   the teleoperation framework to the scenario of wall-following.  The UAV should be operated laterally along a wall with constant distance and orientation to it at a constant altitude.  In this scheme, the movement along the wall is controlled in open-loop by the human operator by directly setting the tilting angle in direction along the wall. By the end of November, the algorithms for laser-based wall following were still under development.

Parallel to the implementation, we were also exploring the question what a minimum sensor set could look like to enable wall-following using a UAV.  We restricted ourselves to use the accelerometer and gyroscope output from an IMU and visual information from a camera. Based on theoretical considerations, we designed an observer that takes as input only the inertial measurements from the IMU and the optical flow from a camera to obtain the metric distance to the wall as well as the metric velocity along the wall.