Embodied Intelligence: Bionic Robot Controller Integrating Environment Perception, Autonomous Planning, and Motion Control
Yahui Gan, Bo Zhang, Jiawei Shao, Zao Han, Ang Li, Xianzhong Dai
Abstract
This letter proposes a bionic robot controller equipped with intelligent perception and autonomous planning modules to address the manufacturing industry's requirements for small-batch, customized, and autonomous task. Three crucial components: motion control module, vision perception module, and autonomous planning module are integrated into the controller based on the ROS framework and ECI platform. Taking multi-object rearrangement problem as an example, a dual-robot collaborative system is established for validation of the controller. The controller deploys the YOLOv5_OBB network for object recognition and localization. Leveraging the task sequence planning and RRT-Growth-Angle algorithm improved in this letter, it autonomously plans collision-free trajectories for the robots, facilitating their movement from the starting point to the grasping position and further to the placement location. The motion control algorithm collaboratively controls dual-robot, ensuring the precise and stable rearrangement of multiple objects into predefined positions. The results affirm that the bionic robot controller effectively mimics the three essential components of human-like functionality—perceiving, pondering, and acting—enabling it to autonomously and intelligently tackle complex tasks, which verifies the viability of the method.