• Use of tightly synchronized hardware modules working together to achieve precision and robustness.
• Vision inputs enhance inertial attitude estimation and provide inputs for map building.
• Inertial inputs ensure robustness in the presence of external interference and noise.
• SLAM methods allow trajectory optimization and visual recognition and reuse of previously visited locations (loop closure).

„Hawkeye" visual inertial navigation module design specs.

• Combine inertial data integration, visual feature matching, depth extraction and pose graph construction support bundle adjustment of keyframe images running with up to high refresh rate (up to 200Hz)
• Use of field proven, industry standard Extended Kalman Filter packed with quaternion representation and a variety of compensation algorithms ensures high stability, precision and scalability
• Ensure centimeter level position estimate and sub-degree attitude solution

When a new path is needed, there are multiple options to follow:

• search based methods (A*)
• optimization based methods (gradient descent optimizations)
• classic geometry (Dubins)

• Employ the most advanced software data fusion and vision processing libraries to be easily integrated with arbitrary new sensors
• Leverage the experience from benchmarking and integrating numerous visual-inertial and LIDAR based SLAM algorithms

• Mapping, trajectory generation and collision avoidance
• computing platform ensures support for hardware-accelerated deep learning modules for further extensibility