Uploaded May 2018 | Updated September 2026, 3 weeks ago
ICRA 2018 Spotlight Video
Interactive Session Tue PM Pod B.5
Authors: Wijayasinghe, Indika; Saadatzi, Mohammad Nasser; ABUBAKAR, SHAMSUDEEN; Popa, Dan
Title: A Study on Optimal Placement of Accelerometers for Pose Estimation of a Robot Arm
Abstract:
This study investigates the effects of inertial sensor placement and noise characteristics on the accuracy of robot pose estimation. Of course, most robots are equipped with joint angle encoders for pose estimation and end-effector positioning. However, in some situations, it's not possible or not desirable to introduce encoders on all joints. Such common examples include legged locomotion, dual arm co-manipulation, and prosthetic limbs. To tackle such situations, one solution is to embed inertial measurement units (IMUs) into artificial skin patches placed on robots' limbs and body. This work analyzes the effects of design parameters such as the number of sensors, their placement on the robot, and noise properties on the quality of robot pose estimation and its signal-to-noise Ratio (SNR). We study the benefits of using a large number of IMUs, which is possible due to the proliferation of inexpensive micro-machined sensors. We use Monte-Carlo simulations and experiments with a two-link robot arm to obtain the distributions of expected estimation error metric values for several accelerometer configurations, which are then compared to determine the optimal number and placement for the IMUs. Results show that the placement of at least two accelerometers on each link has the most significant impact on the pose estimation error, while using a larger number of accelerometers plays a less significant role in reducing the arm pose estimation error and resultant SNR.
ICRA 2018 Spotlight Video
Interactive Session Tue PM Pod B.5
Authors: Wijayasinghe, Indika; Saadatzi, Mohammad Nasser; ABUBAKAR, SHAMSUDEEN; Popa, Dan
Title: A Study on Optimal Placement of Accelerometers for Pose Estimation of a Robot Arm
Abstract:
This study investigates the effects of inertial sensor placement and noise characteristics on the accuracy of robot pose estimation. Of course, most robots are equipped with joint angle encoders for pose estimation and end-effector positioning. However, in some situations, it's not possible or not desirable to introduce encoders on all joints. Such common examples include legged locomotion, dual arm co-manipulation, and prosthetic limbs. To tackle such situations, one solution is to embed inertial measurement units (IMUs) into artificial skin patches placed on robots' limbs and body. This work analyzes the effects of design parameters such as the number of sensors, their placement on the robot, and noise properties on the quality of robot pose estimation and its signal-to-noise Ratio (SNR). We study the benefits of using a large number of IMUs, which is possible due to the proliferation of inexpensive micro-machined sensors. We use Monte-Carlo simulations and experiments with a two-link robot arm to obtain the distributions of expected estimation error metric values for several accelerometer configurations, which are then compared to determine the optimal number and placement for the IMUs. Results show that the placement of at least two accelerometers on each link has the most significant impact on the pose estimation error, while using a larger number of accelerometers plays a less significant role in reducing the arm pose estimation error and resultant SNR.
