| Summary: | Robotics holds the promise of transforming industries, from automating recycling to managing household chores, by enabling machines to perform tasks with human-like dexterity. However, current robotic manipulation systems struggle to achieve the real-time responsiveness required for such tasks. Traditional systems rely on cameras, which slow down control loops with dense and difficult-to-process data. This thesis addresses the need for real-time control in robotic manipulation by utilizing proximity sensors in a high-bandwidth, low-latency object avoidance reflex controller on the Biomimetic Robotics Lab’s dexterous robotic manipulation platform. The research focuses on the two most viable proximity sensors for robotic manipulation: the STMicroelectronics VL6180X Time-of-Flight sensor and the Thinker Phase-Modulated-Light sensor. These sensors are characterized based on their measurement range, error, variance, field-of-view, and convergence time to determine their usability in an object avoidance reflex. Following characterization, a study on the integration of these sensors into the manipulation platform is performed to assess sensing latency and bandwidth implications. Finally, validation of the optimal sensor-controller configuration for the object avoidance reflex—averaging two time-of-flight sensors with a linear virtual force—shows an improvement in bandwidth from 33 Hz to 115 Hz, enhancing the reactivity and stability of the object avoidance reflex. Overall, this research provides a comprehensive study on the individual sensor and sensor-integration levels of proximity sensors for object avoidance reflexes. It enables future researchers to be confident in the manipulation platform’s performance for further controls-level research.
|
|---|