Autonomous Unmanned Vehicle Systems
The Institute of Systems Optimization (ITE) is researching on the fusion of sensors with complementary characteristics in integrated navigation systems. The resulting navigation solution can be found in different areas of application. One of these applications is the AirQuad. It is an autonomous Unmanned Aerial Vehicle (UAV), which is developed at the ITE. The focus is in the field of system design, system architecture and the system intelligence.
The UAV is an electrically-powered four-rotor helicopter. Due to its diameter of about 1m and a total weight of about 1.5 kg the UAV is classified into the class of micro aerial vehicles (MAV). Owing to its ability to hover, it is suited very well for many rescue and security applications.
The focus of the ITE lies on the development of an UAV that is capable to autonomously accomplish exploration missions. Possible areas of application are for example buildings or industrial estates after building fire, industrial accidents or natural disasters. Inertial navigation and control systems for autonomous flights commonly require GPS information to limit inertial sensor errors. In the aforementioned situation, GPS is not always available or the information is impacted due to multipath and shading. Therefore, complementary sensor information (laser rangefinder's scans, camera images, barometer-altimeter measurements and magnetometer measurements are used to aid an inertial navigation system. Beside the self-localization an environment recognition is crucial for autonomous systems. Thus, critical situations (e.g. collisions with walls) can be prevented and passageways into buildings can be detected. This allows precise flights through passageways into buildings for indoor exploration.
A quadrotor with rotors which can be tilted adds a new flexibility in the usage of this class of MAVs. This results in more degrees of freedom in divided attitude and velocity control as well as in reaction time. Basically, the quadrotor design is similar to a common quadrotor. The difference is given by the opportunity to tilt every propulsion motor around the side arm, on which it is mounted. While a common quadrotor has a direct dependency between attitude and velocity in the system itself, the quadrotor with tiltable rotors can perform attitude and velocity changes independently of each other. The system is an over actuated one, due to the fact that there are more control inputs than degrees of freedom. Using a simple control algorithm will fail on this problem because ambiguities exist. Thus, the ITE is researching nonlinear controlling algorithms the quadrocopter with tiltable rotors and two extensions to the control loop in order to increase the overall controller performance.
Localization in Urban Areas
In the vicinity of buildings, GNSS signals are subject to blockage and multipath. Hence, the navigation solution of GNSS-aided navigation filters deteriorates in urban areas. In order to reduce these errors, the ITE develops methods for improving navigation systems by fusing different sensors. Moreover, by adapting navigation filters to different scenarios and implementing integrity checks, availability and reliability are increased significantly. This allows for autonomous flights of MAVs in urban areas, which is a major research objective.
Autonomous indoor exploration
The aerial investigation of buildings can provide rescue forces with information beyond their own field of vision and thus reduce risks. Hereby unmanned Micro Aerial Vehicles (MAVs) lend themselves to perform supporting reconnaissance flights. To do so the agent has to independently explore accessible interior space, react flexibly to obstacles and return safely to the ground station. The MAV, limited in both computing and loading capacity, is faced with a multitude of dynamic challenges which classical systems often cannot meet in real time. That is why at the ITE a novel multi-sensor system for autonomous exploration in GNSS-denied environment is developed. It comprises reactive strategies for secure and effective indoor application: on-board computation of heterogeneous sensor data is ought to enable not only comprehensive reconnaissance but both situational awareness and robustness during the navigation. Data acquisition tailored to mission specific tasks simplifies and accelerates e.g. the generation of plant layouts for rescue forces.
Using various sensors, particularly laser rangefinders, the agent is able to construct a metric map of its surroundings. Based on that, path planning algorithms for navigation can be applied. However, metric maps do not provide further information about the robot's operating environment, which is necessary for both robot and human to understand the scenery. Therefore, the metric map is augmented with mission-specific, qualitative features. As a result, the task-planning capabilities of rescue teams and the situation detection capabilities of aerial vehicles are improved.