Ergodic Exploration Using Binary Sensing for Nonparametric Shape Estimation

Abstract

Current methods to estimate object shape—using either vision or touch—generally depend on high-resolution sensing. Here, we exploit ergodic exploration to demonstrate successful shape estimation when using a low-resolution binary contact sensor. The measurement model is posed as a collision-based tactile measurement, and classification methods are used to discriminate between shape boundary regions in the search space. Posterior likelihood estimates of the measurement model help the system actively seek out regions where the binary sensor is most likely to return informative measurements. Results show successful shape estimation of various objects as well as the ability to identify multiple objects in an environment. Interestingly, it is shown that ergodic exploration utilizes non-contact motion to gather significant information about shape. The algorithm is extended in three dimensions in simulation and we present two dimensional experimental results using the Rethink Baxter robot.

Data-Driven Measurement Models for Active Localization in Sparse Environments

“…We compare our method using ergodic control with another method that is similar in nature known as expected entropy reduction (EER) [1,32,33]. The algorithm is similar to that used in [11,14,23,24] where a location in the search space is chosen based on where the expected change in entropy is maximized.…”

Discrimination of Object Curvature Based on a Sparse Tactile Sensor Array

“…A more common method of curvature discrimination research is that the manipulator moves along the contour of the object, and the surface curvature is solved by the spatial trajectory information of the movement [12][13][14][15][16][17][18]. Back in 1996, M. Charlebois et al [12,13] planned the exploratory movement of a mechanical tentacle on the surface through contact perception.…”

“…In 2007, J. Tian et al [16] rebuilt the surface by tracking along three concurrent curves on the surface to obtain curvature information. In 2017, Ian Abraham et al [18] used a low-resolution binary contact sensor for ergodic exploration and successfully obtained curvature information.…”

Conan Veitch

“…The Mathew Mezić met-ric is widely used by researchers in single and multi-robot research due to being relatively computationally inexpensive. [29,[31][32][33][34]. [14] also provides an algorithm for creating trajectories for multi-agent systems called Spectral Multiscale Coverage (SMC).…”

“…Each robot can have different dynamics if necessary, though each robot will attempt to explore the closest area of high information density within the distribution as long as their control laws are informed by the metric. This applies to any trajectory generation method using this metric, though trajectory generation itself may put restrictions on the agents dynamics [8,14,33]. The system will attempt to match the spatial-average of the domains distribution with the time-average of all of the robots.…”