Neural Network Accelerated Force Estimation Rate for RMIS

This paper presents a neural network approach for upsampling force data in vision-based force estimation for robot-assisted minimally invasive surgery (RMIS). This interpolation enables sampling frequencies upwards of 500Hz, dramatically increased compared to modern vision-based estimation methods alone. With greater sampling frequency, force estimation can support direct force feedback options for RMIS (compared to current, indirect methods, e.g. visual or auditory), which may increase situational awareness and thus surgical performance. The proposed network is optimized and evaluated for its viability in two ablation studies with promising results, achieving sampling rates upwards of five times the accepted minimum rate of 500Hz for direct force feedback and absolute errors within maximum force granularity of 0.06N.