Embedded Files
Introduction
Introduction
The efficient disassembly of end-of-life electric vehicle batteries(EOL-EVBs) is crucial for green manufacturing and sustainable development. The current pre-programmed disassembly conducted by the Autonomous Mobile Manipulator Robot(AMMR) struggles to meet the disassembly requirements in dynamic environments, complex scenarios, and unstructured processes.
In this paper, we propose a Battery Disassembly AMMR(BEAM-1) system based on NeuralSymbolic AI.
It detects the environmental state by leveraging a combination of multi-sensors and neural predicates and then translates this information into a quasi-symbolic space. In real-time, it identifies the optimal sequence of action primitives through LLM-heuristic tree search, ensuring high-precision execution of these primitives. Additionally, it employs positional speculative sampling using intuitive networks and achieves the disassembly of various bolt types with a meticulously designed end-effector.
Importantly, BEAM-1 is a continuously learning embodied intelligence system capable of subjective reasoning like a human, and possessing intuition. A large number of real scene experiments have proved that it can autonomously perceive, decide, and execute to complete the continuous disassembly of bolts in multiple, multi-category, and complex situations, with a success rate of 98.78%.
This research attempts to use NeuroSymbolic AI to give robots real autonomous reasoning, planning, and learning capabilities. BEAM-1 realizes the revolution of battery disassembly. Its framework can be easily ported to any robotic system to realize different application scenarios, which provides a ground-breaking idea for the design and implementation of future embodied intelligent robotic systems.
Project
Project
The whole system includes four levels: hardware, execution, task, and motion. They are the indispensable parts that make up an AMMR, and also the core innovations covered by our solutions to the various challenges in battery disassembly.
Neural Predicate
Neural Predicate
To achieve a more intelligent system, we introduce neural predicates to help BEAM-1 for environment state recognition based on the Neural Symbolic AI. Each neural predicate can be regarded as a neural network, which maps the multi-sensor perception information of the environment to the quasi-symbolic space to complete the characterization of the state. Neural predicates can be arbitrarily combined to describe the current complex state more accurately.
Action Primitive
Action Primitive
Having accomplished the precise perception of the environment state, we realize high-precision control based on action primitives at the execution level. We subdivided the disassembly process and defined 12 action primitives such as Approach, Mate, Push, Insert, and so on. Each primitive is defined by Planning Domain Define Language with execution pre-requirements and execution target effects in symbol space, which will be used for searching during task planning. The definition of primitives ensures that BEAM-1 can autonomously plan appropriate action sequences in dynamic and complex environments to cope with various environmental states and accomplish various tasks.
Video
Video
This video introduces and shows the function and performance of BEAM-1 robot in a real circumstance to accomplish its disassembly task. Click to start the video.
BEAM-1-Demo.mp4Contact us
Contact us
If you have any question, please contact any of us by the following e-mail.
mingchen@sjtu.edu.cn
zhi.gang.wang@intel.com
Page updated
Google Sites
Report abuse