The integration of computer-based technologies interacting with industrial machines or home appliances through an interconnected network, for teleoperation, workflow control, switching to autonomous mode, or collecting data automatically using a variety of sensors, is known as Internet of Things (IoT). When applied inside an industrial context, it is possible to immediately benefit from the analytics obtained, contributing to process optimization, machine health, the safety of workers and asset management. IoT can assist real-time platforms in remotely monitoring and operating a complex production system with minimal intervention of humans. Hence it can be beneficial for hazardous industries, such as mining, by increasing the safety of personnel and equipment while reducing operation costs. An ideal smart automated mine could potentially be achievable by gradually taking advantage of IoT. Currently, different sensors are used in mine-related activities, such as geophones in exploration and blast control, piezometers in dewatering and toxic gas detectors in working frontlines. However, a fully integrated automated system is challenging in practice due to infrastructural limitations in communication, data management and storage. Moreover, the tendency of mining companies to continue with traditional methods instead of relying on untested novel techniques decelerates this progress. In this study, the adaptability of the mining industry to IoT systems and its current development is reviewed. Significant challenges of this progress are investigated and recommendations to develop a comprehensive model suited for different mining sections such as exploration, operation and safety considering flexible technologies such as Wireless Sensor Networks and the introduction of Global Data Management.

This paper presents an algorithm for optimizing the scheduling of trackless equipment in underground mines. With the shortest working interval and maximum productivity as goals, a genetic algorithm (GA) is used to solve the problem, and obtain the optimal working sequence with the most suitable equipment configuration possible. The input for the proposed method is the number of units and capacity of trackless equipment, the production process, ore amount in stopes, and the distance between stopes. The algorithm is verified using four setups of 5 stopes with 5 cycles, 5 stopes with 15 cycles, 10 stopes with 10 cycles, and 10 stopes with 30 cycles. The solution time of the algorithm is no more than 20 min, which is acceptable for practical applications. The results show that the setup of 10 stopes with 30 cycles is closer to the actual production of the mines, and the optimization model can effectively improve the operation efficiency. In this scenario, the robustness of the optimization is tested by simulating equipment failure events. Under the condition of 8% failure rate, the operation time is extended over 3.21-14.56% than expected, which represents strong robustness. The algorithm can quickly provide a feasible and effective solution for the production scheduling decision of trackless equipment in underground mines.

Value functions are central to Dynamic Programming and Reinforcement Learning but their exact estimation suffers from the curse of dimensionality, challenging the development of practical value-function (VF) estimation algorithms. Several approaches have been proposed to overcome this issue, from non-parametric schemes that aggregate states or actions to parametric approximations of state and action VFs via, e.g., linear estimators or deep neural networks. Relevantly, several high-dimensional state problems can be well-approximated by an intrinsic low-rank structure. Motivated by this and leveraging results from low-rank optimization, this paper proposes different stochastic algorithms to estimate a low-rank factorization of the Q(s, a) matrix. This is a non-parametric alternative to VF approximation that dramatically reduces the computational and sample complexities relative to classical Q-learning methods that estimate Q(s, a) separately for each state-action pair.

No abstract