AI-Facilitated Robots

Artificial-Intelligence-Enabled Adaptive Robots

Introduction: The recent development of industrial manufacturing and social services has witnessed a significant trend of automation and intelligentization due to the wide application of robots and the technology of artificial intelligence (AI). While robots liberate humans from tedious and dangerous work in hazardous environments, AI simplifies the programming of robots by automatically inferring patterns and models from the interaction between the robots and the environment. Nevertheless, the application of robots and AI to more general manufacturing and social tasks is still limited by the lack of flexibility and adaptability to the changes in the task and the environment. For example, the conventional way of programming a robot is platform-specific, which means that the control algorithms must be developed for the specific robots individually. In this sense, reprogramming is needed if another robot is assigned to the same task or this robot is required to perform a new task. Even for learning-based approaches, massive training, and data collection are necessary to develop control policies for specific robots. Transplanting or migrating algorithms or policies to new robots or new tasks are challenging problems. Thus, a new concept, adaptive robotics, has been proposed to address the desire that an AI-facilitated robot should be able to properly reprogram itself to these changes without human intervention. Nevertheless, this concept is yet too abstract to provide any specific guidance to the development of robot programs. In this research, we are dedicated to proposing novel concepts, frameworks, and approaches to define and prescribe the connotations and methodologies of adaptive robots. More specifically, we investigate how to use AI-based methods, such as machine learning, reinforcement learning, transfer learning, imitation learning, and meta-learning to facilitate the capabilities of robotic systems in variable environments and changeable tasks. The scope of this research also includes how to upgrade a robot control algorithm to complicated environments and tasks from simpler ones without using massive data, or how to migrate a well-trained algorithm in a simulation environment to reality incorporating the existence of sim-to-real gap. Based on this, we aim to lead the way toward a new generation of robotic devices that are able to automatically adapt themselves to new environments and new tasks with the least possible data and retraining. The main objectives of this research are summarized as follows.

1. Review the existing studies in the literature and define adaptive robots from conceptual and methodological perspectives.

2. Develop high-fidelity simulation environments and experimental platforms for the studies of adaptive robots.

3. Propose novel approaches to cope with the sim-to-real challenge for reinforcement learning methods.

4. Develop effective technologies for data-efficient learning approaches using data aggregation and exploitation.

Period: 2021.03 - 2023.02

Project: Artificial Intelligence enabled highly Adaptive Robots for Aerospace industry 4.0 (AIARA).

Publications:

• Z. Zhang, J. Hong, A. M. S. Enayati, H. Najjaran*, "Using Implicit Behavior Cloning and Dynamic Movement Primitive to Facilitate Reinforcement Learning for Robot Motion Planning". (Submitted) [arXiv]

• A. M. S. Enayati, Z. Zhang, K. Gupta, and H. Najjaran*, "Exploiting Symmetry and Heuristic Demonstrations in Off-policy Reinforcement Learning for Robotic Manipulation". (Submitted) [arXiv]

• J. Hong, Z. Zhang, A. M. S. Enayati, and H. Najjaran*, "Human-Robot Skill Transfer with Enhanced Compliance via Dynamic Movement Primitives". (Submitted) [arXiv]

• R. Dershan, A. M. S. Enayati, Z. Zhang, D. Richert, and H. Najjaran*, "Facilitating Sim-to-real by Intrinsic Stochasticity of Real-Time Simulation in Reinforcement Learning for Robot Manipulation", in IEEE Transactions on Artificial Intelligence, Early Access, doi: 10.1109/TAI.2023.3299252. [IEEEXplore][arXiv]

• A. M. Soufi Enayati, Z. Zhang, and H. Najjaran*,  "A methodical interpretation of adaptive robotics: Study and reformulation", in Neurocomputing, Vol. 512, no. 2022, pp. 381-397, 2022,  doi: 10.1016/j.neucom.2022.09.114. [ScienceDirect][ResearchGate]

• Z. Zhang, D. Wollherr, and H. Najjaran*,  "Disturbance estimation for robotic systems using continuous integral sliding mode observer", in International Journal of Robust and Nonlinear Control, Vol. 32, no. 14, pp. 7946-7966, Sept. 2022, doi: 10.1002/rnc.6252. [Wiley]

• Z. Zhang, R. Dershan, A. M. S. Enayati, M. Yaghoubi, D. Richert, and H. Najjaran*, "A High-Fidelity Simulation Platform For Industrial Manufacturing by Incorporating Robotic Dynamics Into an Industrial Simulation Tool", in IEEE Robotics and Automation Letters, Vol. 7, no. 4, pp. 9123-9128, 2022, doi: 10.1109/LRA.2022.3190096. [IEEEXplore]