Robotic Surgery UI Tracking Correction With LQE Up-Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robotic surgery systems face challenges with noisy or imprecise tracking input data from sensor circuits, which can lead to inaccuracies in the movement of surgical tools, and often have sampling rates that do not match the requirements of the control system, resulting in potential latency issues.
Innovation Solution
A linear quadratic estimator (LQE) is employed to correct and up-sample the tracking input data, integrating latency reduction and up-sampling functions within its prediction phase, using a Kalman filter gain to compute updated output samples and reduce latency by tuning a latency coefficient, and calculating velocity terms through fitting lines to previous output samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the sensor circuit sampling rate is increased to match the controller sample rate, then the tracking input precision is improved, but the latency and computational complexity increase
Solution Approach 1:
The LQE performs preliminary estimation and prediction of the UID position and orientation between actual sensor samples. The prediction phase uses the system model to forecast state variables at intermediate time points, effectively pre-computing values that would otherwise require higher sampling rates. This preliminary action reduces the need for high-speed sampling while maintaining precision.
Solution Approach 2:
The LQE acts as an intermediary between the low-rate sensor circuit and the high-rate controller. It processes the noisy tracking input through its estimation algorithm, producing smoothed and predicted output samples that bridge the sampling rate gap. The Kalman filter gain computed in the update phase mediates between measurement noise and prediction accuracy, delivering clean data to the controller without requiring the sensor to operate at controller speeds.
2Speed
If the LQE prediction phase runs at a higher rate to reduce latency, then the responsiveness is improved, but the computational load increases
Solution Approach 1:
The LQE dynamically adjusts its operation between two phases: prediction phase running at higher rates for responsiveness, and update phase triggered by actual sensor samples. The system model parameters and Kalman filter gains are dynamically computed based on current state estimates and measurement quality. This dynamic switching allows the computational load to be distributed efficiently, with intensive prediction calculations performed only when needed for responsiveness.
Solution Approach 2:
The LQE employs periodic update cycles synchronized with the sensor sampling rate, interspersed with intermediate prediction steps. The prediction phase repeats at higher rates periodically, while the update phase occurs at the lower sensor sampling rate. This periodic structure allows the system to maintain high responsiveness during prediction intervals while reducing overall computational load by limiting full updates to necessary moments.
3Measurement precision
If the tracking input is filtered to reduce noise, then the measurement precision is improved, but the latency increases due to smoothing delays
Solution Approach 1:
The LQE uses feedback through the Kalman filter gain to adaptively weight the balance between predicted values and actual measurements. When measurement noise is high, the gain reduces reliance on noisy inputs; when measurements are reliable, it increases weighting. This feedback mechanism achieves noise filtering without the fixed delay characteristics of traditional smoothing filters, as the estimation adapts in real-time based on actual data quality.
Solution Approach 2:
The LQE changes its internal parameters dynamically through the Kalman filter gain computation. The gain value adjusts the effective filtering strength based on the ratio of measurement noise to prediction uncertainty. This parameter adaptation allows the system to achieve optimal noise reduction at each time step without introducing fixed latency delays, unlike conventional filters with predetermined smoothing constants.
Data Source
AI summary
A sequence of input samples that are measures of position or orientation of an input device being held by a user are received. A current output sample of a state of linear quadratic estimator, LQE, is computed that is an estimate of the position or orientation of the input device. The current output sample is computed based on i) a previously computed output sample, and ii) a velocity term. An updated output sample of the state of the LQE is computed, based on i) a previously computed output sample, and ii) a new input sample. Other embodiments are also described and claimed.