Model based estimation of ellipsoidal object using artificial electric sense

Abstract

In this article we address the issue of shape estimation using electric sense inspired by the active electric fish. These fish can perceive their environment by measuring the perturbations in a self-generated electric field caused by nearby objects. The approach proceeded in three stages. Firstly the object was detected and its electric properties (insulator or conductor) identified. Secondly, the object was localized using the MUSIC (MUltiple SIgnal Classification) algorithm, which was originally developed to localize a radio wave emitter using a network of antennas. Thirdly, the shape estimation relied on the concept of generalized polarization tensor (GPT), which enabled modeling the electric response of an object polarized by an ambient electric field. We describe the implementation of the approach through numerous experiments. The system was able to estimate shape with an average error of 16%, and opened the way toward further improvements. In particular, self aligning the sensor with the ellipsoid through a reactive feedback makes the shape estimation errors drop to 10%.

A Purely Model-Based Approach to Object Pose and Size Estimation With Electric Sense

“…In a second phase, some maps built beforehand in simulation are used to determine the ellipsoid aspect ratio and the object's lateral distance and size using an equivalent sphere model and a Gaussian process regression. The second contribution was proposed by Lanneau et al [13]. Similarly, the location and size estimation of ellipsoidal objects were separated and solved in cascade.…”

“…However, our experiments have shown that the model remains reliable for an object whose size is below 1/4 of the robot length (5cm). Finally, in our experiments, we consider that the orientation θ of the object ranges in [0, π] and the object conductivity γ o is either equal to 1e 5 for conductive objects, or 1e −5 for insulating ones [13]. The parameter space being bounded, when a perturbation on the electric measurements is detected, our method consists in discretizing the object parameters interval, to create a base of candidate solutions, and to test each of them to find the best fitting between estimations and measurements (the lower g).…”

“…4.b), depending on the sign of δI ax,1 we can infer on which side (left or right) of the sensor, the object is, along with its electric properties γ o . There are 4 possible scenarios detailed in [13]. Then, we assign the intervals of the location of the equivalent sphere such as:…”

“…For each experiment, the computational time under Matlab for the whole trajectory was about 80s (in comparison the experiment duration was about 60s). To quantify the estimation results, we used the 3 following metric errors of [13] defined by:…”

Underwater pre-touch based on artificial electric sense

“…In (Servagent et al 2013), the current measurement mode has been implemented through a family of sensors consisting of some plastic slender probes on which are arranged a few metal electrodes in different configurations (see one of them in figure 1-b). In recent years, all these sensor technologies were used to address different issues in underwater robotics ranging from model-based localization (Solberg et al 2008, to reactive navigation (Boyer et al 2013), underwater docking (Boyer et al 2015), and shape recognition (Bai et al 2015, Lanneau et al 2017. Recently, the electric sense has been proposed as an assistance for visual feedback in virtual reality (Fang et al 2016).…”