Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Multi-Sensor Fusion in Continuous-Time
The aim of this project is to gain further introspective into suited approaches to tackle tight integration of individual sensors within a multi-modal suite. Specifically, it concerns augmenting and implementing means of tightly coupling sensory information, originating from a set of Inerti
Keywords: Continuous-time Simultaneous Localization and Mapping, Cumulative Cubic B-Splines, Multi-Sensor Fusion, Tight Sensor Coupling, Direct IMU Integration
CTSLAM is the task of maneuvering in an, a priori, unknown space while mapping its environment and simultaneously estimating the relative transformation of the moving reference frame in said map. Within the context of mobile robotics, gaining an accurate and useful understanding of a robot's surroundings is elementary to successfully achieve any autonomous tasks, such as navigation or obstacle avoidance. Despite aforementioned task representing one of the most well-studied problems in robotics, approaches based on continuous-time formalisms have only been marginally explored thus far. This despite their huge potential to significantly reduced the complexity and struggle to fuse multi-rate sensors in a multi-modal setup. Specifically, within the context of standard frame-based SLAM algorithms, IMU measurements are usually pre-integrated to a common timestamp (i.e. to the time associated with arriving camera frames). However, with the advent of event-based sensors, the paradigm of frame-based SLAM starts crumbling due to complete asynchronicity of sensory inputs. Leveraging the benefits of B-Spline-based continuous-time representations said asynchronicity can be addressed in a principled manner. In particular, given a C2-continuous trajectory estimation, one is also able to infer the instantaneous angular velocity and linear acceleration of the system at any possible point in time. Observing this circumstance, we strive to exploit the information inherent to B-Splines to integrate asynchronous IMU measurements, originating from multiple sensors, in a straight-forward manner into the underling trajectory optimization and to quantify the gains in accuracy of the trajectory estimation by making use of multiple IMUs.
CTSLAM is the task of maneuvering in an, a priori, unknown space while mapping its environment and simultaneously estimating the relative transformation of the moving reference frame in said map. Within the context of mobile robotics, gaining an accurate and useful understanding of a robot's surroundings is elementary to successfully achieve any autonomous tasks, such as navigation or obstacle avoidance. Despite aforementioned task representing one of the most well-studied problems in robotics, approaches based on continuous-time formalisms have only been marginally explored thus far. This despite their huge potential to significantly reduced the complexity and struggle to fuse multi-rate sensors in a multi-modal setup. Specifically, within the context of standard frame-based SLAM algorithms, IMU measurements are usually pre-integrated to a common timestamp (i.e. to the time associated with arriving camera frames). However, with the advent of event-based sensors, the paradigm of frame-based SLAM starts crumbling due to complete asynchronicity of sensory inputs. Leveraging the benefits of B-Spline-based continuous-time representations said asynchronicity can be addressed in a principled manner. In particular, given a C2-continuous trajectory estimation, one is also able to infer the instantaneous angular velocity and linear acceleration of the system at any possible point in time. Observing this circumstance, we strive to exploit the information inherent to B-Splines to integrate asynchronous IMU measurements, originating from multiple sensors, in a straight-forward manner into the underling trajectory optimization and to quantify the gains in accuracy of the trajectory estimation by making use of multiple IMUs.
- Literature review and familiarization with IMUs, Splines and Continuous-Time SLAM.
- Conceptualization/augmentation of existing code to multiple IMUs in Matlab.
- Evaluation against existing approaches and single IMUs setup.
- (Optional) Integration into and testing within a larger CTSLAM framework in C++.
- Literature review and familiarization with IMUs, Splines and Continuous-Time SLAM. - Conceptualization/augmentation of existing code to multiple IMUs in Matlab. - Evaluation against existing approaches and single IMUs setup. - (Optional) Integration into and testing within a larger CTSLAM framework in C++.
- Solid mathematical background and maturity with Matlab and/or C/C++.
- Prior knowledge about Simultaneous Localization and Mapping, IMUs respectively, is beneficial.
- Some familiarity with Linux, ROS, Eigen and Ceres is desirable.
- Solid mathematical background and maturity with Matlab and/or C/C++. - Prior knowledge about Simultaneous Localization and Mapping, IMUs respectively, is beneficial. - Some familiarity with Linux, ROS, Eigen and Ceres is desirable.
David Hug (dhug@ethz.ch) and Marco Karrer (marco.karrer@ethz.ch)
David Hug (dhug@ethz.ch) and Marco Karrer (marco.karrer@ethz.ch)