A collaborative robot (cobot) is a robot designed for direct interaction with humans within a specified collaborative area. Colgate, Edward, Peshkin, and Wannasuphoprasit initially introduced the concept in 1996, starting with a basic version featuring a single joint, often referred to as a steerable wheel. Subsequently, diverse iterations of cobots with varying embedded technologies have entered the market.
Image credit: Gumpanat/Shutterstock
The International Federation of Robotics (IFR) reports continuous growth in the cobot market, with end-users and engineers exploring optimal configurations of sensors, grippers, and user-friendly programming interfaces for effective integration in manufacturing.
Cobots are primarily intended for shared workspaces alongside human counterparts, eliminating the need for physical barriers and facilitating tasks with varying degrees of interaction. These interactions encompass coexistence, synchronization, cooperation, and collaboration. Coexistence denotes proximity between the worker and robot without workspace sharing. Synchronized interaction involves shared workspace, but not concurrently; tasks are completed sequentially. Cooperative interaction sees direct contact between worker and robot, but they tackle different tasks. Finally, collaboration interaction entails direct contact and joint work on the same tasks by both the worker and the robot.
Differences between Cobots and Robots
Cobots and robots share similarities in their capacity to automate tasks traditionally performed by humans. However, they differ significantly in their roles. Cobots work alongside human operators, enhancing productivity through increased force, accuracy, and intelligence. They facilitate faster learning and simpler programming with artificial intelligence (AI) assistance. In contrast, industrial robots replace human workers and typically require complex reprogramming by skilled engineers. Cobots offer real-time interactive interfaces, while robots rely on remote interaction.
The Growing Demand for Cobots
The surge in Industry 4.0 tech adoption within manufacturing and the demand for tailored products have escalated the need for robots. Traditional industrial robots fall short, being less adaptable, expensive, and time-intensive. Cobots offer a solution by collaborating with humans, eliminating these drawbacks.
Cobots find a prime market in small and medium-sized enterprises (SMEs), comprising approximately 90 percent of global businesses, crucial for economic growth and job creation. Due to their affordability, safety, and plug-and-play functionality, they readily integrate into SMEs, yielding swift returns on investment (ROI) while enhancing productivity across the board.
The evolution of Industry 4.0 technologies drives the imperative for cobot deployment in manufacturing, propelled by factors such as market globalization, shorter product lifecycles, heightened customization, labor dynamics, demographic shifts, agility demands, digital transformation, and more.
They excel in tasks such as picking, packing, quality assurance, and more across various industries. Their adaptability makes them valuable in electronics, aerospace, automotive, and other sectors. They also find applications in agriculture, healthcare, and surveillance. Cobots are poised to evolve with advancing AI technology, continuing to perform precise and sophisticated tasks.
Nonetheless, challenges persist in developing robots for real-world applications, particularly in autonomous decision-making and cognitive awareness. Integrating AI and machine learning shows promise across diverse fields. Cobots enhance productivity and safety in settings such as test laboratories, where full autonomy may not be feasible due to lower test volumes.
Cobot-related Work with AI
Machine learning, a subset of AI, leverages algorithmic and statistical processes to enable automated learning from experience. It enhances assembly processes, reducing maintenance and inspection costs and improving quality control post-assembly. It enables non-destructive examination, predictive maintenance, and optimized supply chain management.
However, machine learning's effectiveness depends on vast and relevant data. Subcategories such as deep learning have gained prominence due to increased computing power. Robots, empowered by machine learning, make intelligent, secure decisions, transforming industries beyond manufacturing and warehousing. Researchers have explored various approaches, such as neural networks and reinforcement learning, to create flexible, natural, and adaptive human-robot collaborations (HRCs).
Several studies involving cobots encountered challenges in performing collaborative tasks due to safety concerns. For non-collaborative tasks, AI was applied in different works, such as control algorithms and linear mixed effects models. For example, Bagheri et al. introduced a bidirectional and transparent interaction-based learning system between humans and cobots. This learning system enhances performance through a transparent graphical user interface (T-GUI). This approach allowed the cobot to explain its actions, and operators could provide necessary instructions. It was confirmed that providing explanations improved performance in terms of effectiveness and efficiency.
Amarillo et al. proposed an industrial conventional robot with cooperative learning that enhances the admission controller algorithms for robotic-assisted spine surgery. Similarly, Nicora et al. presented a control framework to support workers collaborating with cobots while maintaining a positive psychological state. Their MindBot coworker combined cobots and Avatar, an interactive virtual system, enhancing working conditions through gaze, gestures, and speech interactions. While another study detailed a modeling approach for robotic devices and a reinforcement learning robot that autonomously adjusted its behavior based on human activity, Story et al. explored how robot velocity and proximity settings affected workload and trust during HRC activities. They found that a robotic system could impact workload even if compliant with safety regulations.
Many studies have applied AI, including deep learning and reinforcement learning, to design cobots for various industrial tasks. For instance, Zhang et al. optimized task assignment in assembly operations using an HRC-reinforcement learning system, and Silva et al. enabled direct control of a robot through video streams from cameras. Additionally, De Winter et al. employed interactive reinforcement learning to transfer skills between humans and robots for cost reduction and accelerated learning, and Ali Ghadirzadeh et al. developed a human-centered collaborative robotic system using reinforcement learning and graph convolutional networks. Akkaladevi et al. introduced a reinforcement learning framework for collaborative assembly processes, while other studies designed a collision detection framework for industrial cobots using deep neural networks and employed deep reinforcement learning for pick-and-place tasks. Furthermore, Chen et al. improved HRC activities with neural learning-enhanced admission control, Qureshi et al. introduced intrinsic motivational reinforcement learning for human-robot interaction (HRI), and Wang et al. used deep learning for continuous human movement observation and prediction in HRC.
These studies collectively demonstrate the diverse applications and advancements in collaborative and industrial robots, incorporating AI and machine learning to improve safety, efficiency, and HRI.
Future of Cobots
Existing research predominantly focuses on specific, fixed tasks for robots, lacking an in-depth examination of dynamic scenarios involving diverse tasks. Current collaboration techniques rely on static, feature-based methods or human-controlled robots, with human supervisors making most decisions. This limits the cognitive capabilities of service-providing robots. Establishing effective cooperation between robots and human workers for safe and efficient task execution can be challenging, as any incorrect or missed commands may lead to task failure, necessitating human intervention.
Deep learning techniques currently in use lack real-time efficiency and adaptability for cobots in complex situations where real-time application is not feasible. Advancements are needed in online deep learning for distributed teams of cobots and human operators, facilitating data-driven control system advancements. Current technologies do not sufficiently consider sharing knowledge and experiences among cobots to enhance learning. Achieving this requires unique distributed sensor signal processing and data aggregation across wirelessly networked robots.
Based on current developments in AI and robotics, the landscape of cobot research is broadly divided into economic, social, and technological dimensions.
Economic Dimensions: The global cobot market's growth has attracted major players, influencing cobot types and markets. These markets encompass geographic regions and various industrial sectors. Cobots are poised for integration in healthcare, education, and manufacturing, but proving their value in new settings is vital. Smaller SMEs may face challenges if their original plans shift.
Social Dimensions: Safety is paramount, reflecting Isaac Asimov's 'Three Laws of Robotics.' Recent safety standards address co-working scenarios. Human factors, ergonomic concerns, and labor market inclusion must be considered, along with collaboration with workers, as tasks evolve to involve computational devices.
Technological Dimensions: Research focuses on simplifying cobot programming and enabling advanced interactions with semantic understanding and AI-aided anticipation. Programming tasks shift to shop floor operators, requiring simplicity and intuitiveness.
Achieving advanced interactions involves capturing human cues, distinguishing between intentional and non-intentional gestures, and developing semantic understanding. Predicting human intent, future actions, and movements is essential. The technological frontier adapts machines to humans in the evolving landscape of human-machine interaction.
References and Further Readings
Kakade, S., Patle, B., and Umbarkar, A. (2023). Applications of collaborative robots in agile manufacturing: a review. Robotic Systems and Applications, 3(1), 59–83. DOI: https://doi.org/10.21595/rsa.2023.23238
Knudsen, M., and Kaivo-Oja, J. (2020). Collaborative Robots: Frontiers of Current Literature. Journal of Intelligent Systems: Theory and Applications, 3(2), 13–20. DOI: https://doi.org/10.38016/jista.682479
Borboni, et al. (2023). The Expanding Role of Artificial Intelligence in Collaborative Robots for Industrial Applications: A Systematic Review of Recent Works. Machines 11, no. 1: 111. DOI: https://doi.org/10.3390/machines11010111
Sajwan, M., Singh, S. (2023). A Review on the Effectiveness of Machine Learning and Deep Learning Algorithms for Collaborative Robot. Archives of Computational Methods in Engineering 30, 3489–3508. DOI: https://doi.org/10.1007/s11831-023-09903-2