Profil #88348

The project Entern is concerned with creating technologies for partial autonomous operation of robots in lunar and planetary exploration missions. Specifically it covers technologies for the robust autonomous exploration of craters and caves in a context of space missions.

These types of environments are of special interest to the scientific community and provide potential places for future infrastructures and habitats due to their protected locations.

My contribution addresses the localisation issues a robotic system faces when navigating and mapping a structurally symmetrical, and relatively featureless environment such as the ones encountered.

Helping build a portable embedded platform for Visual Inertial Navigation using Extended Kalman Filter (EKF) and Simultaneous Localization and Mapping (SLAM) techniques. The project uses GPS-less sensor data from a monocular camera, accelerometer and an odometer. My work responsibilities include

Transferring and customizing BLAS optimised algorithms to GPU using libraries like Magma, cuSPARSE, CLapack and cuBLAS.

Feature extraction in C++ and CUDA combination running on NVIDIA GPU.

Sensor fusion using Extended Kalman Filter (EKF) approaches.

CUDA programming on NVIDIA Tegra X1 board.

Optimization of code segments which use large sparse matrices.

Outlier feature detection and depth estimation using triangulation and baseline optimization.

Code profiling using tools like gprof, valgrind, zoom, oprofile and perftools.

Porting code to and from C++ and Python.

Embedding Python code in C++ using custom made Python/C API wrappers.

Extending Python with C++ code.

Project code assimilation, debugging and compiling.

Writing makefiles and automating compilation for medium sized project.

• Documentation in form of changelogs, status reports, references etc.