Autonomous Systems, Robotics and Machine tools

Lead Researcher
Alexandre Simões
UNESP

Associate Researchers
Alexandre da Silva Simões
UNESP
Anarosa Alves Franco Brandao
POLI/USP
Antonio Cesar Germano Martins
UNESP
Carlos Cesar Aparecido Eguti
ITA
Carlos Henrique Costa Ribeiro
ITA
Carlos Henrique Quartucci Forster
ITA
Denis Silva Loubach
ITA
Douglas Bellomo Cavalcante
IPT
Emilia Villani
ITA
Esther Luna Colombini
UNICAMP
Flávio Alessandro Serrão Gonçalves
UNESP
Helmo Kelis Morales Paredes
UNESP
Jaime Simão Sichman
POLI/USP
Jefferson de Oliveira Gomes
IPT
Leliane Nunes de Barros
IME/USP
Leopoldo André Dutra Lusquinho Filho
UNESP
Luis Armando de Oro Arenas
UNESP
Manuel Antonio Pires Castanho
IPT
Marcos Ricardo Omena de Albuquerque Maximo
ITA
Maria Glória Caño de Andrade
ITA
Rodrigo Bueno Otto
FPTI-BR
Ronnie Rodrigo Rego
ITA
Sandra Lúcia de Moraes
IPT
Silvestre Eduardo Rocha Ribeiro Junior
---

Lines of Research
Autonomous Systems, Robotics and Machine tools

This research line focuses on various stages, ranging from conceptual research (TRL 03) aimed at enhancing system autonomy to more advanced stages of technological maturity (TRL 06), where prototyped versions of these technologies are developed. These technologies are inherently integrated with electrical, mechanical, and computational dimensions, and are primarily characterized as intelligent robots and machine tools for industrial applications.

Autonomy in an AI system refers to its intrinsic ability to make decisions, that is, to act without human intervention (RUSSELL; NORVIG, 2020). Therefore, these systems are typically capable of:

  1. Perceiving the environment they are immersed in, using various sensors and multiple communication channels characteristic of Industry 4.0 and the Internet of Things (IoT), generating a vast amount of data.
  2. Manipulating this data (developing models, conducting analyses, recognizing patterns, making selections, grouping, querying, deriving, diagnosing, forecasting, or even learning) to constitute and/or expand their knowledge about the world.
  3. Occasionally, planning their actions proactively in the environment, considering the allocation of agents and/or resources in time and/or space, with the aim of optimizing their objectives.
  4. Acting in real-time, if necessary, in the environment, enabling various gains in comparison to conventional industrial processes.

In this scenario, machine learning gains particular relevance, which is the ability to provide machines with systematic approaches that allow them to improve their performance in a specific task through successive experiences (ALPAYDIN, 2014). In its most applied form, this line aims to develop specific solutions for the needs of different industrial segments, ensuring their transition to new levels of production, quality, and efficiency (e.g., energy efficiency). Within this research area, the following themes stand out:

  • Intelligent Systems: This research line focuses on the study and development of complex systems aimed at industrial problems. Its field of investigation includes areas such as: Model-Based Systems (e.g., inference in logical or rule-based models, state model searching, heuristics, evolutionary or fuzzy models, Bayesian networks, Markov decision processes, etc.); Automated Planning (deterministic, probabilistic, continuous, or hybrid); Scheduling; Distributed and Multi-agent Systems; Cognitive models (CIORTEA; MAYER; MICHAHELLES, 2018; MARSLAND, 2015);
  • Machine Learning: This research line focuses on the development and enhancement of techniques to improve machine performance through experience (data). Its field of investigation includes techniques such as: Supervised Learning (regressions, SVMs, Artificial Neural Networks, Deep Learning, Decision Trees, Bayesian Learning); Unsupervised Learning (k-means, k-medoids, SOMs, Vector Quantization, Mixture Models, Expectation-Maximization, LDA, PCA); Reinforcement Learning; Deep Reinforcement Learning; Evolutionary Learning; Imitation Learning; Classification; Clustering; Dimensionality Reduction (CIORTEA; MAYER; MICHAHELLES, 2018; GOODFELLOW; BENGIO; COURVILLE, 2016; SUTTON; BARTO, 2018);
  • Intelligent Automation: This research line is aimed at solving applied problems through the proposition of integrated solutions in electrical, mechanical, and computational domains. Its field of investigation includes areas such as: Machine Vision; Pattern Recognition; Sensing, Localization, and 3D Mapping; Intelligent Instrumentation; Automated Diagnostics; Automated Inspection; Consumption Estimation; Fatigue and Lifespan Prediction; Smart and Reconfigurable Machines and Tools; Internet of Things (IoT); Smart Manufacturing; Human-Machine Collaboration; Productive Process Optimization (BMWI, 2019; DOPICO et al., 2016); Energy Efficiency; Smart Energy Grids; Demand Response; Optimization (FRASER; ANASTASELOS; RAVIKUMAR, 2018; IEEE, 2018; MAKONIN; POPOWICH; GILL, 2013; SOUZA et al., 2018; TALARI et al., 2017; ZHANG; HUANG; BOMPARD, 2018); Manipulator Robots; Cooperative and Collaborative Automation; Service Robots; Wheeled Robots; Legged Robots; Aerial Robots (drones); Aquatic Robots (SIEGWART; NOURBAKHSH; SCARAMUZZA, 2011).

Some of the major challenges of AI in this domain are: (i) dealing with sensorial and data multimodality; (ii) transforming vast amounts of information into useful knowledge about the world; (iii) adapting AI techniques to the context of each industrial segment; (iv) operating in distributed environments; (v) keeping up with the evolution of computational techniques; (vi) generating integrated solutions in electrical, mechanical, and computational dimensions; (vii) integrating machines and humans working together in the same tasks and environments. Furthermore, more flexible and dynamic automated systems can be modeled as sociotechnical systems where agents, whether humans or machines, manage one or more industrial devices (IoT artifacts with embedded intelligence) and/or control independent sets of variables, coordinating to achieve objectives and optimize system performance.

The application of AI techniques can have the following potential impacts: (i) increased productivity; (ii) improved product quality; (iii) enhanced fault detection; (iv) reduced production, operation, and maintenance costs; (v) increased yield; (vi) improved workplace safety; (vii) decreased setup and maintenance time; (vii) flexibility in the production process; (viii) expanded and integrated information about products and processes; (ix) cooperation between different machines and between humans and machines; (x) fostering the development of new products and business models.

 

References