Embedded Files
Project Overview
Project Overview
- A highly modular generic data fusion system
- Synthesize tracking data from an uncertain number of asynchronous sensors in the presence of uncertainties
- The Extended Kalman Filter, Monte Carlo Method, and the principle of Back Propagation
- Feedback mechanism for control decisions
- Cross-platform development including MATLAB and ROS-centered embedded system
Video Demo
Video Demo
The Problem of Sensors
The Problem of Sensors
In actual scenarios, sensor observation will be influenced by various kinds of factors. There might be noises, sudden errors or steady shifts during the robot operation. Relying on any single sensor cannot always guarantee a correct perception of outside environment.
For example, the inertia measurement unit (IMU) observation may have noises and accumulated error as shown in the figure below.
Simulation Result
Simulation Result
The simulation shows that even though some sensors are having errors throughout the whole process and some sources provide huge sudden error, the final fusion result still keeps on the right track.
Cross-Platform Application
Cross-Platform Application
The algorithm was initially realized in visualized MATLAB environment to accelerate the system development. After the simulation verification, embedded C code is generated, modified, and downloaded to embedded boards.
High-performance Teensy® 3.2 and 3.5 boards are connected as fusion center:
ROS system is used as the core unit and exchange data among sensors and fusion processing boards. And the modularity allows the extension to more sensors.
Final Operating Result
Final Operating Result
Google Sites
Report abuse