Current advances and trends in the fields of mechanical, material, and software engineering have allowed mining technology to undergo a significant transformation. Aiming to maximize the efficiency and safety of the mining process, several enabling technologies, such as the recent advances in artificial intelligence, IoT, sensor fusion, computational modeling, and advanced robotics, are being progressively adopted in mining machine manufacturing while replacing conventional parts and approaches that used to be the norm in the rock ore extraction industry. This article aims to provide an overview of research trends and state-of-the-art technologies in face exploration and drilling operations in order to define the vision toward the realization of fully autonomous mining exploration machines of the future, capable of operating without any external infrastructure. As the trend of mining at large depths is increasing and as the re-opening of abandoned mines is gaining more interest, near-to-face mining exploration approaches for identifying new ore bodies need to undergo significant revision. This article aims to contribute to future developments in the use of fully autonomous and cooperative smaller mining exploration machines. Rock ores have an integral role in product manufacturing. It is widely accepted that modern society’s needs and ore extraction requirements are highly correlated. It is without a doubt that traditional methods, which require the participation of human personnel, will not be sufficient to satisfy production demand, since they have a high downtime and may cause harm to the personnel in hazardous environments [ 1, 2]. It is also important to consider that the inefficiency of human-curated processes may cause harm to the environment. Lastly, it is becoming increasingly more important to reduce the overall environmental impact, whenever possible [ 3, 4]. The aforementioned problems underline a need for a radical transformation in the way the mining industry undergoes exploration and production drilling, as well as extraction processes. This transformation can be observed through an already developing trend, as there is an increase in automation and the use of robots in these fields. Regarding software implementations, from human-assisted monitoring systems to remotely operated or fully autonomous mining operations, the boundaries of automation have substantially expanded within the mining engineering field. This article reviews automation in drilling and mining, focusing on trends and state-of-the-art technologies for face exploration and production drilling. In other words, this review focuses on fully autonomous, smaller, and cooperative mining machines. The mapping of knowledge domains for automated operations in drilling and mining referenced in this article can be examined in Table 1. The rest of this article is structured as follows. Section 2 will briefly describe the current state of the mining industry, focusing on the current challenges that it faces and how fully automated mining can be the main route to overcoming those challenges. Section 3 will provide an overview of drilling approaches, focusing on conventional methods used in the mining industry. Section 4 will present the state of the art in automated mining and intelligent systems, summarizing current advancements and research trends. In Section 5, the technological infrastructure utilized in automated mining will be presented, including various sensors and software crucial to these processes. Section 6 will touch upon potential advancements in intelligent mining systems, such as the integration of artificial intelligence (AI) and neural networks, aiming to enhance safety, efficiency, and precision in automated operations. Finally, Section 7 will conclude the article by sharing additional research visions, particularly emphasizing the development of infrastructure-less systems adaptable to various deep mine and abandoned mine situations. Assessing the current state of the mining industry, it can be observed that an important part of the operations is conducted sub-optimally, while the need for resource extraction is constantly increasing. In the meantime, the quality of ore grades located on the surface is decreasing, which leads to the extraction of ores from deeper depths [ 69]. Furthermore, the increasing focus on net-zero emission policies encourages mining companies to adopt more environmentally friendly policies that require the acquisition of different equipment and raises the need for more effective ways of mining [ 66, 76, 77, 78, 79]. Society’s interest in worker health and safety also requires companies to protect their workers from extreme mining conditions, such as dust, dirt, poorly lit areas, and areas with a high potential for rock tumbling [ 78]. The current issues that the rock ore mining industry is facing can be resolved or at least mitigated by fully automating the processes of mining and drilling while utilizing intelligent systems [ 77, 79]. Intelligent systems can increase efficiency if they are used for the scheduling and monitoring of the operation of fully autonomous machines such as hauling fleets. This efficiency can be translated into a reduction in machine wait time, a reduction in equipment energy consumption and resource usage, and the ability to adapt to random events that would normally cause disruptions in the ore extraction process [ 78]. Additionally, fully automating the mining industry can aid in the protection of workers, who can observe machines operating in hazardous areas such as blasting zones with the use of proper sensors from virtually identical environments (DTs) [ 77]. The fully automated and intelligent approaches discussed in this paper cannot be implemented immediately. There are plenty of specifications that need to be met in order for the proposed technologies to be mature enough to be implemented in raw material production. For instance, the intelligent methods that are currently being researched are characterized by high computational intensity and increasing learning time depending on the project’s complexity [ 66]. Additionally, both fully automated processes and intelligent systems need to be robust, flexible, and adaptable, incorporating uncertainties that are a result of insufficient data or unpredictable events that might occur during the mining operation [ 78]. Furthermore, for intelligent systems to be implemented, maintenance and operational costs must be considered, along with the large variety of smart sensors needed for the collection of proper data. Finally, the reliability of these sensors under difficult conditions, such as environments with high dust concentration, humidity, and a lack of proper lighting, should be considered [ 77]. Drilling has been the subject of significant change throughout history, providing many different approaches when it comes to mining. To date, the approaches described below have appeared in the mining industry. Cable tool drilling was the first developed drilling technique, and it was used to extract salt from salt quarries [ 5]. Its modern implementation follows the same principle: a drilling bit is lifted through a mechanism to a great height and is then allowed to fall down the borehole to break the rock formation [ 5, 6], and after that, a bailer collects the fragments [ 6]. Auger drilling utilizes a screw-like drill bit to break the formation [ 5]. This method does not usually include the use of mud; instead, it uses the helical path of its bit, called a flight, to force the fragments to the surface [ 6, 7]. Rotary drilling is a technique that utilizes the rotational motion of the drill bit to deepen the borehole [ 5]. It uses a motor and a long shaft to rotate the drill bit and a mining fluid, called mud, to force the fragments to the surface, and it then sieves through the fluid to separate the solid materials [ 6]. Diamond drilling applies the same principle of operation that the rotary drilling technique uses. The main difference between the two is that diamond drilling relies on the abrasive wear its drill bit causes to produce an uninterrupted cylindrical output of whatever ore it is drilling. This is achievable because its bit is actually a metallic matrix with embedded synthetic industrial diamonds with high hardness [ 5, 6]. Directional drilling allows the boring procedure to be performed at a controlled angle and normally features a down-the-hole motor. The system may or may not have the ability to change its angle multiple times [ 8]. Normally, this can be accomplished in a multitude of ways, such as the following: Whipstock: A special geometry that allows the drill bit to slide off its surface. The geometry’s angle is what directs the drilling bit [ 8]. Bent-house PDM: The motor’s housing is at an angle, allowing the motor to drill at a specific, predefined trajectory [ 8]. Best sub: In this configuration, there are hubs that connect two different bent subs at different angles. It provides a greater degree of freedom when drilling [ 8]. Whipstock: A special geometry that allows the drill bit to slide off its surface. The geometry’s angle is what directs the drilling bit [ 8]. Bent-house PDM: The motor’s housing is at an angle, allowing the motor to drill at a specific, predefined trajectory [ 8]. Best sub: In this configuration, there are hubs that connect two different bent subs at different angles. It provides a greater degree of freedom when drilling [ 8]. The evolution of directional drilling systems led to Rotary Steerable Drilling Systems ( RSDSs). These systems can change the drill bit’s trajectory mid-operation, allowing the implementation of complex trajectory planning and exploration, as reported in [ 9, 10]. These systems are primarily down-the-hole motor systems that can be divided based on the method they employ to affect the bit’s direction and can be roughly divided into the following types: Push-the-bit systems: Push-the-bit systems have a pump that pushes actuators fixated on the system’s external housing. These pneumatic systems utilize the polar array of these actuators, normally pads, to push the bit to the desired direction [ 10], as can be seen in Figure 1. The most famous implementations are the Schlumberger PowerDrive and Baker-Hughes Autotrak systems, as in [ 9, 11]. Point-the-bit systems: These systems vary from the push-the-bit systems in that the actuators reside on the inside of the configuration [ 10, 11]. The most well-explored example of this is the Halliburton Sperry-sun Geo-Pilot system [ 9]. In these cases, the control of the direction of the bit is possible by using two eccentric rings, changing its angle directly using motors, as shown in Figure 2. Hybrid systems: Hybrid systems are, in essence, a combination of the previous two systems. They have multiple actuators that assist them in better controlling every aspect of the drilling procedure. In the Schlumberger PowerDrive Archer system, the bit’s shaft uses a universal joint and internal actuators to push to orient it at the desired angle [ 9, 11]. Push-the-bit systems: Push-the-bit systems have a pump that pushes actuators fixated on the system’s external housing. These pneumatic systems utilize the polar array of these actuators, normally pads, to push the bit to the desired direction [ 10], as can be seen in Figure 1. The most famous implementations are the Schlumberger PowerDrive and Baker-Hughes Autotrak systems, as in [ 9, 11]. Point-the-bit systems: These systems vary from the push-the-bit systems in that the actuators reside on the inside of the configuration [ 10, 11]. The most well-explored example of this is the Halliburton Sperry-sun Geo-Pilot system [ 9]. In these cases, the control of the direction of the bit is possible by using two eccentric rings, changing its angle directly using motors, as shown in Figure 2. Hybrid systems: Hybrid systems are, in essence, a combination of the previous two systems. They have multiple actuators that assist them in better controlling every aspect of the drilling procedure. In the Schlumberger PowerDrive Archer system, the bit’s shaft uses a universal joint and internal actuators to push to orient it at the desired angle [ 9, 11]. Longwall systems ( Figure 3) are large robotic systems that combine the drilling and hauling operations through the use of a conveyor belt system and several horizontal motors with drill bits that perform the drilling [ 12, 13]. Famur’s MIKRUS longwall system features a cutting and loading head, which has two cutting drums that can break coal and a conveyor belt that can retrieve it from the mine [ 14]. It also has a multitude of sensors that allow the process to be even more automated [ 15]. Automated hauling is the process in which the drilled formations are reclaimed automatically. The state-of-the-art choice for such a process is a robotic hauler, also called a scraper. Scrapers are intelligent systems that can transport ore fragments from the drilling point to a preset point. They utilize sensors to understand their position and orientation in the world and use weight sensors while digging the ore fragments to provide data about the extracted ore [ 16]. Additionally, they can be operated autonomously or remotely by a human. Such a system is shown in Figure 4. These robots are designed with a dual capability to handle various challenges in complex environments. They can operate autonomously, adjusting to issues like limited signal propagation in underground areas, or be controlled by human operators when direct guidance is needed. Aside from scrapers, there are also automatic mining trucks that are used for the same purpose as the scrapers, but they have a much larger capacity than the scrapers and are used to retrieve extracted ore [ 16]. In mining, a controlled explosion can be used to break the rock formation. This process may prove dangerous due to the possible cave-in of the mine. Instead of using humans to transport and charge explosives, robots are used to carry, secure, and charge them [ 16, 17]. Rock drilling is a process that is implemented during excavation procedures or during the production of ore through blasting. Depending on the method of drilling, different machines are used. Such machines are the hydraulic rock-drilling jumbo and the down-the-hole (DTH) drill. Intelligent and unmanned versions of these machines allow for efficient ore mining and accurate blasting. This efficiency stems from systems that incorporate smart blockage prevention mechanisms for the drilling pipes, alongside a fully automated drill-pipe bank that oversees the management of the drilling tubes based on the drill’s wear status. The accuracy in blasting is supported by systems that utilize anti-deviation control technology that controls the drilling pipes in real time [ 16]. By observing the trends mentioned in this on the automation process of mining operations, it can be noticed that there are different automated systems that can partially fulfill the tasks needed for mining without the need for a human presence on site. Given the fact that the collapse of rocks in mines is one of the common causes of fatal accidents [ 1], it seems that removing human workers from these sites can lead to an important increase in the safety of the overall mining industry. Aside from the sensors used by specific robotic systems to perform tasks, there are sensors that orchestrate the coordinated action of several systems and provide analytics regarding the mining progress. Laser-Induced Breakdown Spectroscopy is a method of examining the contents of the core extracted by a mining machine. It utilizes a laser to emit photons onto the cylinder’s surface, which creates plasma that excites electrons, and a spectrometer measures the response photons. This technology was recently adapted as an integrated system in down-the-hole equipment after being exclusively used in the laboratory [ 18, 19, 20]. Lidar technology has always been an important technology when the mapping of an area is required. There are currently systems that employ this technology to map and navigate entire caves and mines through the application of SLAM algorithms [ 21, 22, 23, 24], such as automatic loading trucks [ 25]. An example of the sensor’s output is shown in Figure 5. A time-of-flight camera (TFC) or sensor (TFS) uses human-made signals of light, which can be created by either a laser or an LED [ 26, 27]. The laser TFC is a scanner-less LiDAR type of sensor. The main difference that separates these sensors from LiDARs is that only one pulse is needed to capture the surrounding scene instead of many consecutive scans. Real-world applications of TFCs show their practicality. Such an example is their use in the local planning of a group of quadrupedal robots for the exploration of mines, as described in [ 51]. Sensors provide useful data that can later be utilized for the localization of the robot and the workers. Successful localization and sensor information require a robust and effective network to transmit the information to all the machines and workers, both in the mine and at the control center. Currently, there are navigation platforms that allow for the precise localization of each robot in a 3D-digitized environment [ 28]. This can be implemented by using multiple navigation stations with lasers mounted on top of them, which allow them to perceive their pose through rotary encoders and predefined installation location coordinates [ 16] or through alternative advanced positioning methods [ 29]. It is evident that localization in the mining environment may prove difficult due to the fact that technologies and methods, such as GPS and motion-based sensor tracking, may prove inaccurate or unavailable in deep mines. High-end systems often employ additional robots whose job is to place easily recognizable constructs so that the robots can extrapolate their positions [ 30, 31]. In order to solve the problem that the underground environment poses during the localization process of robotic systems and personnel, non-RF methods have been employed, which include the use of magnetic induction methods [ 32]. Whether they use the Earth’s magnetic field, which varies because of the mine’s unique local structure, or the superposition effect of artificial magnetic fields through coils [ 21], it is possible for the sensor to extrapolate its position extremely accurately [ 33] if the coils’ positions are known or if the local magnetic field is already mapped. WiFi is a major RF technology that can be implemented in mines in the form of a large network of signal repeaters to help the propagation of information in complex tunnels [ 21, 24]. Another major RF technology is LoRaWAN (Long-Range WAN), which operates at a lower frequency (that translates to a larger wavelength), which enables stations to receive and transmit data through soil and minerals at a greater distance [ 21, 34, 35], albeit with lower bandwidth. Digital twins (DTs) are digitized models of the actual electromechanical elements and mimic their behavior throughout the process. They provide the operators with a more comprehensive visualization of the physical world’s state than traditional graphs by using multiple sensors and reducing resource expenditure while allowing the implementation of more complex and targeted software for decision-making. The cost of implementing DTs comes in the form of increased hardware demands and the need for high computational power, which is reflected by the quality of the sensing systems [ 36]. Ref. [ 37] proposes the six-layer digital twin architecture shown in Figure 6. By combining IoT sensors, cloud networking, and simulation software, it is possible to transform the physical world into a digitized model and apply secure and automated data-driven decision-making to carry out equipment maintenance and control through the different data analysis layers. Furthermore, this model allows streamlined information communication to every member of the mining crew. While the systems’ hardware may be able to perform various tasks, without the direction of specialized software, it would not be possible to automate them. e-Drilling, located in Stavanger, Norway offers a software suite that allows the constant evaluation of the state of the drill and the well’s borehole. It falls under the general digital twin tools that mimic the movement and state of the actual machine through a simulated model. This software also allows for better visualization and control of the process by the workers through its various modules [ 38, 39]. For example, the wellAhead module is used to simulate a digital twin of the well, and the wellSIM module handles the mechanical loads and drilling parameter simulation of the drill itself. The wellAhead module’s digital twin is depicted in Figure 7. Draco is an open-source library that compresses and decompresses three-dimensional geometrical meshes and point clouds [ 42]. By compressing the data, it is possible to upload them wirelessly with restricted available bandwidth from separate actors with limited computational power to a main computational system, which will then only send back motion commands [ 40, 41, 42]. Forestall uses proven predictive algorithms for the mining and oil industries and, coupled with cloud-based machine health and predictive maintenance products, offers optimized asset performance with minimized downtime. Platform-agnostic health monitoring ensures comprehensive insights into critical machinery’s condition [ 43, 44]. TIMining Aware creates real-time mine and block model visualization. The live mine plan compliance feature ensures that operations align with predefined strategies and regulatory requirements, and global accessibility across multiple devices fosters collaboration and productivity among users [ 44, 45]. The above technologies have all been utilized in individual units and in networks of robots. The true extent of their usefulness can only be expressed through the coordination of many different systems for the betterment of the mining process. Currently, in the mining industry, several types of robots are being used. As mentioned above, autonomous hauling trucks as a part of an Autonomous Haulage System (AHS), especially in open-pit mines [ 82], drilling robots, charging robots, and automated conveyor belts are some of the categories that they can be separated into [ 83, 84]. Examples of commercially available collaborative systems include the AutoMine drill and machine fleet, which allow the simultaneous control of several pieces of mining equipment [ 47, 61, 85]. Another example is Caterpillar’s MineStar Command for Underground Systems, which automates multiple hauling trucks while allowing human operators to watch from afar [ 46, 47, 48, 49]. An additional commercial application is EH-RemoteHeadControl v2 by Elgór-Hanses S.A., which requires the collaboration of a roadheader, a power supply unit, and some of the many sensors mentioned above for the creation of a DT around the shearer so that the roadheader can be operated remotely and safely for the creation of drifts in the mine [ 15]. Furthermore, research trends include examining the benefits of automated systems such as AHSs in open pits and exploring how decentralized AHSs can be applied in underground mines [ 86]. Also, the usage of Unmanned Ground Vehicles (UGVs) and Unmanned Aerial Vehicles (UAVs) is considered beneficial for mine inspections and rescue missions [ 50, 51, 52, 53, 54]. Research on systems of collaborative robots such as AHSs shows that they allow more efficient battery and fuel consumption in autonomous vehicles in comparison to manually driven ones while removing the operators of the hauling tracks from the mines and assigning them the role of observers. Furthermore, there is no longer a delay when changing shifts, since the trucks operate continuously [ 82, 86, 87]. Research on the use of UGVs and UAVs for inspection and rescue missions aims to increase the safety of inspectors from dangers that they may face, such as high concentrations of toxic gases or weak mine ceilings that might lead to the need for a rescue mission or even fatal accidents [ 52, 88]. In rescue missions, where time is of the essence, the use of robots aims to reduce the focus on the safety of the rescue team, thus providing trapped miners with a faster and more efficient rescue plan [ 50, 54]. The current infrastructure shows that it is possible to digitize mines and deploy robots that can collaborate on a peer-to-peer basis, while the human factor can be integrated with the role of the observer. Aside from the aforementioned and already implemented systems, there are also numerous others that are currently being tested or researched. Some examples are listed below: Drill piston control for better bit-pointing accuracy: This approach promises better results when it comes to pointing the bit of an RSDS. It uses multiple pistons, valves, water, and stabilization clamp pads to better control the orientation and force applied by the bit [ 55]. Photogrammetry in the mining industry: While Virtual Reality technology is being used today for minor tasks, it is being researched and worked upon extensively due to the presently unexplored advantages described in [ 56, 57], especially with the improvement in modern cameras. Mining professionals are already adopting technologies from research, especially within the area of the digitization and mapping of mines with high-resolution cameras. Additionally, large-area scans for open-pit sites, which once required the use of special fiducial markers (ground control points) [ 58], are no longer needed with the introduction of Real-Time Kinematics and Post-Process Kinematics technologies [ 59]. Furthermore, there is ongoing research on multi-camera systems that can produce rock-mass digital models with a camera rig and can also support UAVs [ 25]. Neural networks for augmented personnel safety: Neural networks can be used for a diversity of tasks. Research has shown that one of their uses can be the monitoring of the activity of personnel inside the mine to decrease accidents and exposure to hazardous areas [ 60]. One such example is the use of a YOLOv8 model ( Figure 8) to determine the location of humans and hazardous areas [ 61]. YOLOv8 is a convolutional neural network that makes predictions of bounding boxes and class probabilities all at once. Virtual Reality simulation for multi-agent system action planning: Virtual Reality’s importance is already underlined, but the range of its applications cannot be stressed enough. Currently, there are ready-to-use software solutions that allow for the simulation of multiple robots in a virtual mining environment, expressed through the game development engine Unity3D [ 62]. Simulators of this kind may prove essential in mining procedure planning, especially in more complex systems with multiple interactions [ 57]. AI-assisted DTs: Advancements that are currently being made in the field of artificial intelligence suggest that AI will be further integrated into FEA simulation software to provide better insights regarding mining data [ 63, 64]. Surveys report that AI will be refined enough to be widely accepted in the industry by 2030 [ 65], providing a better bridge of communication between the real and digital worlds. Optimized haulage scheduling using Reinforcement Learning: Reinforcement Learning (RL) methods can be used to schedule the operation of fleets of autonomous haulers in real time utilizing model-free Q-Learning algorithms. In a simulated environment, this method resulted in a reduction in the wait time of each hauler and reduced the emission of greenhouse gases that resulted from the extensive fuel consumption of the machines. This reduction in fuel consumption was as high as 30% in comparison to fixed-schedule approaches while increasing productivity by 50% [ 66]. Controlling equipment using Deep Reinforcement Learning: Deep Reinforcement Learning (DRL) can be used to control underground, individually operating machines as a substitute for human operators. Machines such as Load-Haul-Dump machines can be controlled with the use of a DRL framework by implementing a soft actor-critic (SAC) algorithm in a simulated environment, lowering energy consumption and increasing productivity [ 68]. In more detail, in [ 68], the energy consumption was found to be 21% lower for the autonomous vehicle in comparison with the energy consumed when a similar vehicle was operated by a human in the real world, and productivity was increased by 7%. Another example of operating multiple machines using RL, implementing proximal policy optimization (PPO) and SAC, can be examined in [ 67]. Using DRL in mine production scheduling: DRL methods can even be utilized for the design of underground mine production layouts [ 69]. Utilizing DRL in the pre-production stages of the mining process allows for a better evaluation of the mineral grade since it utilizes uncertainty to create a stochastic design in comparison to deterministic designs. Deterministic designs are created by the more conventional geostatistical methods, such as mineable shape optimizers, which tend to lead to the under-performance of mining projects. A case study of underground stoping for gold production in [ 69] resulted in 8.3% higher expected profit and 3.4% more gold mined than the baseline that was provided by a mineable shape optimizer. Development and research for environmental monitoring: To ensure that the mining industry does not damage the environment, tools, indices, and methodologies are being developed to ensure the minimization of pollution [ 70, 71]. One such example can be seen in Figure 9. Monitoring of the drilling rig: Research on monitoring drilling bit wear has been conducted to predict and analyze the drilling bit’s condition. These studies have shown that sound and vibration signals, as well as examining the geological and mechanical properties of rocks, can be used to survey the drilling bit’s wear condition [ 72]. Intelligent control of the drilling: Computer vision and Deep Neural Networks for image classification are used to control the drilling rig and improve procedures such as drilling path planning and intelligent hole positioning [ 72]. Measurement while drilling (MWD) and machine learning for rock monitoring: MWD is a technology that provides real-time measurements of drilling parameters, e.g., torque, rotation speed, and drilling speed. These parameters are used to determine the magnitude of the rock stress caused by the drilling operation with the aid of digital panoramic imaging technology for boreholes. Additionally, intelligent rock property sensing can be achieved by implementing machine learning techniques with the data provided by MWD [ 72]. AI and big data technology in ventilation systems: Current research on ventilation systems focuses on acquiring ventilation parameters, such as wind volume and pressure, temperature, humidity, and the detection/concentration of toxic gases. To achieve this, research focuses on the proper placement of the minimum number of sensors that provide reliable data about those parameters. After obtaining these data, diagnostic methods are proposed based on learning algorithms and big data analysis, as well as airflow control methods [ 72]. Intelligent rock stress monitoring: Rock stress during drilling operations causes the deformation of the surrounding rock. Therefore, real-time monitoring methods have been developed based on machine vision and learning algorithms that provide early warning and predictions about surrounding rock deformation and damage [ 72]. Quantum computing for open-pit profile optimization: There are currently studies that show that it is possible to implement quantum computing to determine the best possible parameters for ore extraction in open-pit mines. This could potentially help with the planning of the drill site [ 73, 74, 75]. Drill piston control for better bit-pointing accuracy: This approach promises better results when it comes to pointing the bit of an RSDS. It uses multiple pistons, valves, water, and stabilization clamp pads to better control the orientation and force applied by the bit [ 55]. Photogrammetry in the mining industry: While Virtual Reality technology is being used today for minor tasks, it is being researched and worked upon extensively due to the presently unexplored advantages described in [ 56, 57], especially with the improvement in modern cameras. Mining professionals are already adopting technologies from research, especially within the area of the digitization and mapping of mines with high-resolution cameras. Additionally, large-area scans for open-pit sites, which once required the use of special fiducial markers (ground control points) [ 58], are no longer needed with the introduction of Real-Time Kinematics and Post-Process Kinematics technologies [ 59]. Furthermore, there is ongoing research on multi-camera systems that can produce rock-mass digital models with a camera rig and can also support UAVs [ 25]. Neural networks for augmented personnel safety: Neural networks can be used for a diversity of tasks. Research has shown that one of their uses can be the monitoring of the activity of personnel inside the mine to decrease accidents and exposure to hazardous areas [ 60]. One such example is the use of a YOLOv8 model ( Figure 8) to determine the location of humans and hazardous areas [ 61]. YOLOv8 is a convolutional neural network that makes predictions of bounding boxes and class probabilities all at once. Virtual Reality simulation for multi-agent system action planning: Virtual Reality’s importance is already underlined, but the range of its applications cannot be stressed enough. Currently, there are ready-to-use software solutions that allow for the simulation of multiple robots in a virtual mining environment, expressed through the game development engine Unity3D [ 62]. Simulators of this kind may prove essential in mining procedure planning, especially in more complex systems with multiple interactions [ 57]. AI-assisted DTs: Advancements that are currently being made in the field of artificial intelligence suggest that AI will be further integrated into FEA simulation software to provide better insights regarding mining data [ 63, 64]. Surveys report that AI will be refined enough to be widely accepted in the industry by 2030 [ 65], providing a better bridge of communication between the real and digital worlds. Optimized haulage scheduling using Reinforcement Learning: Reinforcement Learning (RL) methods can be used to schedule the operation of fleets of autonomous haulers in real time utilizing model-free Q-Learning algorithms. In a simulated environment, this method resulted in a reduction in the wait time of each hauler and reduced the emission of greenhouse gases that resulted from the extensive fuel consumption of the machines. This reduction in fuel consumption was as high as 30% in comparison to fixed-schedule approaches while increasing productivity by 50% [ 66]. Controlling equipment using Deep Reinforcement Learning: Deep Reinforcement Learning (DRL) can be used to control underground, individually operating machines as a substitute for human operators. Machines such as Load-Haul-Dump machines can be controlled with the use of a DRL framework by implementing a soft actor-critic (SAC) algorithm in a simulated environment, lowering energy consumption and increasing productivity [ 68]. In more detail, in [ 68], the energy consumption was found to be 21% lower for the autonomous vehicle in comparison with the energy consumed when a similar vehicle was operated by a human in the real world, and productivity was increased by 7%. Another example of operating multiple machines using RL, implementing proximal policy optimization (PPO) and SAC, can be examined in [ 67]. Using DRL in mine production scheduling: DRL methods can even be utilized for the design of underground mine production layouts [ 69]. Utilizing DRL in the pre-production stages of the mining process allows for a better evaluation of the mineral grade since it utilizes uncertainty to create a stochastic design in comparison to deterministic designs. Deterministic designs are created by the more conventional geostatistical methods, such as mineable shape optimizers, which tend to lead to the under-performance of mining projects. A case study of underground stoping for gold production in [ 69] resulted in 8.3% higher expected profit and 3.4% more gold mined than the baseline that was provided by a mineable shape optimizer. Development and research for environmental monitoring: To ensure that the mining industry does not damage the environment, tools, indices, and methodologies are being developed to ensure the minimization of pollution [ 70, 71]. One such example can be seen in Figure 9. Monitoring of the drilling rig: Research on monitoring drilling bit wear has been conducted to predict and analyze the drilling bit’s condition. These studies have shown that sound and vibration signals, as well as examining the geological and mechanical properties of rocks, can be used to survey the drilling bit’s wear condition [ 72]. Intelligent control of the drilling: Computer vision and Deep Neural Networks for image classification are used to control the drilling rig and improve procedures such as drilling path planning and intelligent hole positioning [ 72]. Measurement while drilling (MWD) and machine learning for rock monitoring: MWD is a technology that provides real-time measurements of drilling parameters, e.g., torque, rotation speed, and drilling speed. These parameters are used to determine the magnitude of the rock stress caused by the drilling operation with the aid of digital panoramic imaging technology for boreholes. Additionally, intelligent rock property sensing can be achieved by implementing machine learning techniques with the data provided by MWD [ 72]. AI and big data technology in ventilation systems: Current research on ventilation systems focuses on acquiring ventilation parameters, such as wind volume and pressure, temperature, humidity, and the detection/concentration of toxic gases. To achieve this, research focuses on the proper placement of the minimum number of sensors that provide reliable data about those parameters. After obtaining these data, diagnostic methods are proposed based on learning algorithms and big data analysis, as well as airflow control methods [ 72]. Intelligent rock stress monitoring: Rock stress during drilling operations causes the deformation of the surrounding rock. Therefore, real-time monitoring methods have been developed based on machine vision and learning algorithms that provide early warning and predictions about surrounding rock deformation and damage [ 72]. Quantum computing for open-pit profile optimization: There are currently studies that show that it is possible to implement quantum computing to determine the best possible parameters for ore extraction in open-pit mines. This could potentially help with the planning of the drill site [ 73, 74, 75]. There seems to be a trend of using digital technology to further improve the efficiency and capabilities of mining robotic solutions. This also means that there is a constant need for additional data-gathering sensors and systems that will help tune the function of the software to match the physical needs of the mine. Emerging technologies seem to be heavily dependent on the current state of mining technology, which does not necessarily mean that they will be able to satisfy the current needs of the mining industry. Case studies seem to refer to technologies that follow implementations that require pre-existing communication systems. It is also apparent that modern practices tend to favor non-modular approaches that often rely on specific proprietary equipment and software to operate. While today’s systems seem to be adequate for standard mines, they are not expected to work in environments that have been abandoned before the installation of such infrastructure. While this approach may be sufficient for most mining industries, it is often not flexible enough to cover all of the possible mine configurations. To better generalize automation solutions, creating a unified framework is essential. This framework needs to categorize the different layers and parts of the data handling and the physical aspects of these systems in order to encourage the standardization of the software and methodology used. Additionally, it would be interesting to test rig-less and infrastructure-less methods of automation, since installing these technologies may not be available in old, abandoned mines. For example, it is not realistic to expect that 5G and UHF communication systems will be present in every mine. This points to the fact that decentralized approaches may need to be investigated in a more thorough way. However, the limitations of this approach should be taken into account, such as bandwidth limitations and latency in the communication between different machines and systems. As the mining industry is currently focusing on extraction technologies at large depths and revisiting the extractions at abandoned mines, there are major trends in miniaturizing mining machines, revisiting mining plans, e.g., the volume of drifts, and moving toward a future with completely autonomous mining operations. Even with a technology standard, it is imperative that safety and health considerations be taken into account before sending crew members into mines to install the necessary infrastructure. To that end, it would be beneficial to implement systems that allow for the better inspection of hazardous elements without the necessary control or supervision from a human operator. To achieve this, the introduction of robotic systems that monitor the necessary non-automated operations while also being able to contribute to reducing the personnel’s exposure time to potentially adversarial conditions is envisioned. It is evident that such automation for mining safety, while still an immature field, is something that can be expanded on. Lastly, it seems that while intelligent systems are being studied and developed to reduce environmental strain from mining operations, said systems are not yet mature enough to minimize the environmental impact. Thus, it should be noted that there are still not enough solutions when it comes to automated waste management, and this is something that has to be addressed in the near future. Currently, practices such as the use of waste neutralization tanks and ground storage are used to combat solid mining waste (slag, tailings, etc.), but even they may pose problems with specific types of waste [ 3]. These intelligent systems will require the corresponding machines to operate properly. The machines need to contain adequate sensors and the proper computational power. They require an intelligent and generative design. There are a number of new technologies developed for the automation of the mining process, which is a testament to the ever-increasing need for better efficiency. The inclusion of automation systems is what allows modern manufacturing needs to be met while reducing possible human harm and human error. From partially automated to fully automated systems, this integration reshapes the foundation of the mining industry and engineering. The use of many different types of autonomous and intelligent machines that can communicate on a peer-to-peer basis in situations where pre-existing communication networks are weak, faulty, or nonexistent can raise the efficiency of the ore extraction and increase the safety of the workers in the mines. The use of automated systems for mining operations could improve workers’ health and increase their safety. This might lead to a reduction in needed unspecialized personnel and an increase in personnel with higher educational backgrounds. The digitization of the mining process and the use of more sophisticated digital models can also enhance the remote monitoring of mining operations. Automated mining is still a field with a lot of progress to be made. As time goes by, with continued intellectual and financial investments in intelligent systems design, it will become even clearer that these systems should be embraced and implemented as the new standard in the mining industry. Finally, it should be noted that this article focuses on the automation and robotization of drilling and mining operations, and as such, the focus has been on these areas. Other important aspects of mining, e.g., the effectiveness of the selected methods and environmental impact, have not been addressed, as they are considered out of scope. Conceptualization, A.K. and T.F.; writing—original draft preparation, A.K., T.F. and K.S.; writing—review and editing, A.K., T.F., K.S., G.N. and P.K.; supervision, G.N. and P.K. All authors have read and agreed to the published version of the manuscript. This research work was funded by the Horizon Europe project PERSEPHONE [grant number 101138451]. Data are contained within the article. We would like to express our gratitude toward Achilleas Kousios for providing us with his insights during the editing process of this paper. The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results. The following abbreviations are used in this manuscript: 3D Three-dimensional 5G Fifth-generation mobile network AHS Autonomous Haulage System AI Artificial intelligence DRL Deep Reinforcement Learning DT Digital twin DTH Down the hole FEA Finite Element Analysis GPS Global Positioning System IoT Internet of Things LED Light-Emitting Diode LIBS Laser-Induced Breakdown Spectroscopy LiDAR Light Detection and Ranging LoRaWAN Long-Range Wide-Area Network MWD Measurement while drilling PPO Proximal policy optimization RF Radio Frequency RL Reinforcement Learning RSDS Rotary Steerable Drilling System SAC Soft actor-critic SLAM Simultaneous Localization and Mapping TFC Time-of-flight camera TFS Time-of-flight sensor UAV Unmanned Aerial Vehicle UGV Unmanned Ground Vehicle UHF Ultra-High Frequency YOLO You Only Look Once Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. Abstract Current advances and trends in the fields of mechanical, material, and software engineering have allowed mining technology to undergo a significant transformation. Aiming to maximize the efficiency and safety of the mining process, several enabling technologies, such as the recent advances in artificial intelligence, IoT, sensor fusion, computational modeling, and advanced robotics, are being progressively adopted in mining machine manufacturing while replacing conventional parts and approaches that used to be the norm in the rock ore extraction industry. This article aims to provide an overview of research trends and state-of-the-art technologies in face exploration and drilling operations in order to define the vision toward the realization of fully autonomous mining exploration machines of the future, capable of operating without any external infrastructure. As the trend of mining at large depths is increasing and as the re-opening of abandoned mines is gaining more interest, near-to-face mining exploration approaches for identifying new ore bodies need to undergo significant revision. This article aims to contribute to future developments in the use of fully autonomous and cooperative smaller mining exploration machines. Keywords: drilling robots; mining; mining sensors; autonomous mining