TR2018-087

Evaluation of the Discrete Time Feedback Particle Filter for IMU-Driven Systems Configured on SE(2)*



This paper evaluates the utility of the feedback particle filter (FPF) for state estimation of SE(2)-configured dynamics in a real-time context. The filter is implemented in discrete time to fuse gyroscopic- and accelerometer measurements with Ultra-Wideband (UWB) and camera measurements. With this state information, the FPF is compared to other common filters in terms of the estimate mean square error (MSE) and robustness to initial conditions. An analysis is done on how these metrics scale with utilization of computational resources, concluding that the FPF should be considered for embedded applications with CPUs on par with the Cortex M4 processor.