direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Logo der TU Berlin

Inhalt des Dokuments

Online Inference in Sensor Networks

From Fachgebiet Regelungssysteme TU Berlin

Jump to: navigation, search

These research activities are dedicated to the development of new methods for dynamic inference in a large range of applications that require online information processing for decision making in real time. By combining novel machine learning approaches with well-founded systems and control theory, we aim to provide methods that overcome trade-offs between inference speed and accuracy on the one hand and resilience and explainability on the other hand.

Most research projects are concerned with specific application systems and with leveraging the advantages of the newly devised methods to overcome real-world challenges in autonomous aerial and ground vehicles, collaborative robotic systems and P4 medicine solutions.

Plug-and-Play Gait Analysis

Teaser_Thumb

This project is concerned with the development and evaluation of algorithms that allow an extensive analysis of the human gait based on the measurement data of IMUs attached to the feet, shanks and thighs. Unlike previous results, our approach exploits the geometrical constraints induced by the joint mechanics instead of requiring complex calibration movements or exact sensor mounting.more

Orientation and Joint Angle Estimation in Inhomogeneous Magnetic Fields

Teaser_Thumb

We consider the standard sensor fusion problem in IMU orientation estimation and highlight the malicious effects that magnetic disturbances can have on the inclination portions (roll and pitch) of the orientation estimate. We develop new methods which assure that magnetic field measurements affect only the heading (yaw) component of the orientation estimate.more

Event-Triggered Learning in IMU Networks

Teaser_Thumb

Communication load is a limiting factor in many real-time systems. Event-triggered state estimation and event-triggered learning methods reduce network communication by sending information only when it cannot be adequately predicted based on previously transmitted data.more

Realtime Gait Phase and Terrain Detection

Teaser_Thumb

The gait cycle is modeled by a finite automaton that describes the gait phases for each side. Gait phase transitions are detected in real time using adaptive-threshold-based algorithms. Furthermore, the orientation and position trajectories of both feet are analyzed to detect stairs, slopes and uneven terrain.more

Hand Gesture Tracking and Control

Teaser_Thumb

The adaptive hand neuroprostheses will enable patients with a hand paresis to manipulate objects of the daily live. In order to continuously adjust the stimulation parameters, the hand and finger motion is analyzed in real time by an inertial sensor network that is distributed on the finger and hand segments.more

Motion Tracking in Paraplegic Cycling

Teaser_Thumb

In order to support knee flexion/extension as well as foot dorsiflexion, plantarflexion and eversion, we develop multichannel gait neuroprostheses. The stimulation intensities are adjusted from stride to stride based on inertial sensor measurements and a decentralized iterative learning control scheme.more

3D Visualization and Realtime Biofeedback

Teaser_Thumb

To provide realtime feedback of body posture and motion, the orientations of multiple body segments are estimated based on measurements of a wearable inertial sensor network. Additional information like muscle activity or stimulation intensities are simultaneously visualized on the skin of the avatar.more

Magnetometer-free motion tracking for connected multi-body systems

Teaser_Thumb

To use motion tracking of multi-body systems in indoor environments, methods for orientation estimation without the use of magnetometers are developed. They exploit the kinematic descriptions of the systems to estimate the orientation of each connected object.more

Motion Tracking for Autonomous Search and Rescue Drones

Teaser_Thumb

Autonomous drones that empower search and rescue missions at sea rely on accurate real-time motion tracking. We derive new methods for non-restrictive calibration and sensor fusion that meet the special demands of operating on the high seas.more

Recommend this page