An alternative approach for the acquisition of 6axis force and torque sensor for a Martian robotic arm

This thesis presents an alternative method of data retrieval from a six-axis Force-Torque Sensor (FTS) on a robot arm for Mars surface missions. The context of the research is the Mars Sample Return mission, which involves launching a new rover equipped with an articulated, extended robot arm to collect and carry geological samples initially collected by the Perseverance rover. Accurate force and torque sensing at each of the joints of this robotic arm are critical to perform manipulation accurately, achieve system stability, and ensure collision-free interaction with the Martian environment. The current design incorporates centralized electronics and +5V DC excitation, and therefore analog sensor signals must traverse long distances along flexible PCBs shared among other signal and power lines. This configuration is prone to signal deterioration through electromagnetic interference and cross-talk, especially in the highly populated and electromagnetically noisy environments prevalent in robotic systems. This thesis promotes the use of modulation-based acquisition that shifts the signal into a cleaner frequency before transmission and thus increases immunity to interference. The signal is then demodulated, filtered, and digitized close to the controller. Several key design choices were necessary for developing this method. To begin with, different modulation and demodulation approaches were considered, and square-wave modulation was chosen for its ease of implementation and use with embedded systems. A thorough spectral examination of the signals on the robotic arm's flexible PCB was performed using MATLAB, finding a low-interference region near 7 kHz. This frequency was used as carrier for the modulation, because it lies at a notch of the frequency response of the sigma-delta ADC's internal fifth-order Sinc filter, enhancing noise rejection. The complete acquisition chain was simulated on MATLAB, from generation of the signal to injection of noise, modulation and demodulation circuits, filtering, and digital conversion. Upon verification in simulation, the system was then implemented in hardware on an STM32F401RE microcontroller on a Nucleo development board. The proof of concept includes analog front-end circuits for signal exciting and conditioning, a synchronous demodulation block, and SPI communication readout of the data. The residual noise is further reduced after post-processing in firmware using a digital moving average filter. Experimental results indicated that the proposed method drastically improves signal-to-noise ratio and measurement stability over the traditional DC-based architecture both in nominal and high-interference conditions. The system exhibits high resistance to cross-talk and spectral interference, confirming its viability for accurate sensing in hostile environments. Beyond proposing a sensible enhancement to existing acquisition systems for space robotics, the research presents an extensible generic architecture applicable to other fields where reliable sensor measurements are needed under noisy environments, including industrial control, biomedical systems, and aerospace.