Soft Robot Localization Using Distributed Miniaturized Time-of-Flight Sensors

TL;DR

This work investigates self-localization of a soft robot using distributed miniaturized Time-of-Flight sensors. By converting each sensor's depth map to a point cloud and fusing samples over motion, the method builds a local environment representation and localizes the robot base in a known map using a Particle Filter with ICP-based pose weights relative to the map $\hat{X}$. Experimental results show localization errors around $\approx 0.03$ m in position and $<4.5^{\circ}$ in orientation, with accuracy limited by sensor noise and environmental simplicity. The study demonstrates that distributed $8×8$ ToF measurements can support effective localization on small soft robots, and points to probabilistic sensing models and more feature-rich environments as avenues for enhancement.

Abstract

Thanks to their compliance and adaptability, soft robots can be deployed to perform tasks in constrained or complex environments. In these scenarios, spatial awareness of the surroundings and the ability to localize the robot within the environment represent key aspects. While state-of-the-art localization techniques are well-explored in autonomous vehicles and walking robots, they rely on data retrieved with lidar or depth sensors which are bulky and thus difficult to integrate into small soft robots. Recent developments in miniaturized Time of Flight (ToF) sensors show promise as a small and lightweight alternative to bulky sensors. These sensors can be potentially distributed on the soft robot body, providing multi-point depth data of the surroundings. However, the small spatial resolution and the noisy measurements pose a challenge to the success of state-of-the-art localization algorithms, which are generally applied to much denser and more reliable measurements. In this paper, we enforce distributed VL53L5CX ToF sensors, mount them on the tip of a soft robot, and investigate their usage for self-localization tasks. Experimental results show that the soft robot can effectively be localized with respect to a known map, with an error comparable to the uncertainty on the measures provided by the miniaturized ToF sensors.

Figures (4)

• Figure 1: Soft robot estimating its pose with respect to the environment using a Particle Filter based on measurements collected with miniaturized ToF sensors. The faded robots represent the poses associated with different particles. For each sensor, the measured 8×8 depth map is converted into a point cloud and exploited for localization. The colored points belong to the point cloud associated with the current pose, while the shaded gray dots are the ones previously collected. The robot is successfully localized by using the coarse information retrieved with the ToF sensors.
• Figure 2: The experimental setup is shown. It consists of the cuboid environment (0.7x0.7x0.6 m) considered for the localization task (top left corner), the soft robot with reflective markers rigidly mounted on it (1), three VL53L5CX ToF sensors mounted on the robot tip (2), and the OptiTrack tracking system (3) with four cameras recording the reflective markers poses which constitute the ground truth for the pose estimation.
• Figure 3: The model point cloud $\hat{X}$ is shown with gray dots and over-imposed on the real cuboid used for the localization task. The red point cloud $X_k$, was obtained by merging $k=50$ random point cloud samples recorded by the robot in different poses reached when commanded with $k$ pressure sequences. An exemplification of the soft robot in three different configurations is also shown.
• Figure 4: The bar plot shows the mean linear and rotational error along with their relative standard deviation ranges with a growing number of point cloud samples in two cases: (a) when point cloud samples are merged using the $k$-NN model for the transformation with respect to a common base frame; (b) when point cloud samples are merged using OptiTrack data for the transformation with respect to a common base frame.